Skip to content

Commit

Permalink
add latest tag for release docker image (#279)
Browse files Browse the repository at this point in the history 
Need to tag the latest for the release docker image

---------

Co-authored-by: dhuangnm <[email protected]>
  • Loading branch information
@dhuangnm
dhuangnm and dhuangnm committed Jun 3, 2024
1 parent bd14cbd commit fdd9a31
Show file tree
Hide file tree
Showing 2 changed files with 18 additions and 4 deletions.
9 changes: 7 additions & 2 deletions .github/actions/nm-build-docker/action.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,9 +6,9 @@ inputs:
type: string
required: true
build_type:
description: "type of nm-vllm to install for the docker image: nightly (default) or release"
description: "type of nm-vllm to install for the docker image: NIGHTLY (default) or RELEASE"
type: string
default: 'nightly'
default: 'NIGHTLY'
build_version:
description: "version of nm-vllm to install for the docker image: latest (default) or specific version e.g. 0.4.0, 0.4.0.20240531"
type: string
Expand All @@ -27,6 +27,11 @@ runs:
--build-arg build_type=${{ inputs.build_type }} \
--build-arg build_version=${{ inputs.build_version }} \
--target vllm-openai . || status=$?
if [ ${status} -eq 0 ] && [[ "${build_type}" = "RELEASE" ]]; then
echo "Also tag image for RELEASE build as latest"
docker image tag ghcr.io/neuralmagic/nm-vllm-openai:${{ inputs.docker_tag }} ghcr.io/neuralmagic/nm-vllm-openai:latest || ((status+=$?))
fi
docker image ls -a
echo "status=${status}" >> $GITHUB_OUTPUT
echo "status=${status}"
exit ${status}
Expand Down
13 changes: 11 additions & 2 deletions .github/workflows/publish-docker.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -16,9 +16,9 @@ on:
type: string
default: 'main'
build_type:
description: "type of nm-vllm to install for the docker image: nightly (default) or release"
description: "type of nm-vllm to install for the docker image: NIGHTLY (default) or RELEASE"
type: string
default: 'nightly'
default: 'NIGHTLY'
build_version:
description: "version of nm-vllm to install for the docker image: latest (default) or specific version e.g. 0.4.0, 0.4.0.20240531"
type: string
Expand Down Expand Up @@ -65,3 +65,12 @@ jobs:
target: vllm-openai
push: true
tags: ghcr.io/neuralmagic/nm-vllm-openai:${{ inputs.docker_tag }}

- name: Push image (latest for RELEASE)
uses: docker/build-push-action@v5
if: ${{ inputs.push_to_repository == 'yes' && steps.build.outputs.status == 0 && inputs.build_type == 'RELEASE' }}
with:
context: .
target: vllm-openai
push: true
tags: ghcr.io/neuralmagic/nm-vllm-openai:latest

16 comments on commit fdd9a31

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8724197394622712 prompts/s 3.873385234131611 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3969.230232948828 tokens/s 3970.219864984901 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.25214097261185 prompts/s 17.06325742782891 prompts/s 1.05
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2112.7783264395407 tokens/s 2218.223465617758 tokens/s 1.05
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.775286978017828 prompts/s 5.763501051583438 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2700.2931794420156 tokens/s 2694.7825516783523 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.86803173296103 prompts/s 49.425915520304486 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3241.422062642467 tokens/s 3212.6845088197915 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0440765451876723 prompts/s 2.0402672822810786 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4188.312841089541 tokens/s 4180.50766139393 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7954070169246112 prompts/s 1.801382309359772 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 233.40291220019947 tokens/s 234.17970021677039 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.621898428383183 prompts/s 6.7729068231492615 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 860.8467956898138 tokens/s 880.4778870094041 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9101758890170959 prompts/s 0.9144514774798328 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.32286557222247 tokens/s 118.87869207237826 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46589816388708977 prompts/s 0.4662566113130598 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.98507549058756 tokens/s 126.08200445386888 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.0554887412293 tokens/s 118.14631692932059 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.8877864587622106 prompts/s 2.9294087791973236 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 375.4122396390874 tokens/s 380.8231412956521 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.25523737620285974 prompts/s 0.2569725363697384 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.18085890637177 tokens/s 33.40642972806599 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.469718689691976 prompts/s 4.468047229100103 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1416.40915557649 tokens/s 1415.8794864295314 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1020.6662224160822 tokens/s 1021.09772675701 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.76733156176274 prompts/s 3.7634225711787357 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3861.514850806808 tokens/s 3857.5081354582044 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4483969158389574 prompts/s 2.4478319372827997 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 756.1302581954924 tokens/s 755.9557777512562 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 559.9940780614652 tokens/s 560.8048243833222 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.27767854178434 prompts/s 26.10767148263572 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1708.0491052159823 tokens/s 1696.998646371322 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 48.16555000713796 prompts/s 47.68354528965047 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3130.7607504639673 tokens/s 3099.4304438272807 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9679912020333594 prompts/s 0.9682411901603802 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.632389984967 tokens/s 288.7069306113883 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 210.02827774251813 tokens/s 210.0792910290977 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0433997217330906 prompts/s 2.0394265863400007 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4186.926029831103 tokens/s 4178.785075410661 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8503781247324123 prompts/s 1.8539691473119937 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 587.9218751566458 tokens/s 589.0628531548713 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 414.3169323234179 tokens/s 415.1123439391762 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.891390228124001 prompts/s 7.844609518894351 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4048.2831870276123 tokens/s 4024.284683192802 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.008731772205205 prompts/s 2.001360152379693 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4115.891401248466 tokens/s 4100.786952225991 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.595832587701288 prompts/s 25.564807771210155 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1663.7291182005836 tokens/s 1661.71250512866 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.395833496735422 prompts/s 3.409341889682664 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.4583545756048 tokens/s 443.2144456587463 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.013785641444482 prompts/s 23.67526420447888 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3097.778347746338 tokens/s 3054.109082377775 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839764539862954 prompts/s 0.9839812919439876 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.12873737937247 tokens/s 290.13016386732454 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.4372364941279 tokens/s 212.41860136722823 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.811040345205015 prompts/s 17.724458048825532 prompts/s 1.05
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2185.4352448766517 tokens/s 2304.1795463473195 tokens/s 1.05
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7956364811123715 prompts/s 1.8035648075244035 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 233.4327425446083 tokens/s 234.46342497817247 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.00365779565251 prompts/s 15.836319601059488 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4112.9400534826955 tokens/s 4069.9341374722885 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3958641057077714 prompts/s 3.409011020079548 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.4623337420103 tokens/s 443.17143261034124 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.357019073263755 prompts/s 2.3587430760558656 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 727.9103436658013 tokens/s 728.4427617014128 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 540.9327346219341 tokens/s 541.328390964053 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.356041284435077 prompts/s 5.3909289544752665 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 696.28536697656 tokens/s 700.8207640817847 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.502534600115887 prompts/s 7.492620536692848 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3848.80024985945 tokens/s 3843.714335323431 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.536028822274932 prompts/s 4.542116862078641 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1437.4221734907032 tokens/s 1439.3514124241005 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1036.6640270427129 tokens/s 1038.1765107757747 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9796863767181632 prompts/s 1.9917095136303464 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 257.3592289733612 tokens/s 258.92223677194505 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.931852982277784 prompts/s 13.92541738641096 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3580.4862164453903 tokens/s 3578.8322683076167 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29.309021070704855 prompts/s 28.95521254051672 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3780.8637181209265 tokens/s 3735.2224177266567 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.77542106174842 prompts/s 2.796880131115109 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 360.8047380272946 tokens/s 363.59441704496413 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2395065606129969 prompts/s 0.24061529216630784 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.135852879689597 tokens/s 31.27998798162002 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3942045609256635 prompts/s 3.407268153966697 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.2465929203363 tokens/s 442.9448600156706 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.073674473017117 prompts/s 2.0737548282729463 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4248.958995212073 tokens/s 4249.123643131267 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0123628767454191 prompts/s 1.018507334458996 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 131.60717397690448 tokens/s 132.4059534796695 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.595172433808666 prompts/s 3.5825356745967367 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1139.2741925496282 tokens/s 1135.2697299229599 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 822.1608096347236 tokens/s 819.2709765665965 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.641484802176773 prompts/s 21.103234859219995 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2683.3930242829806 tokens/s 2743.4205316985995 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.640222238599318 prompts/s 10.643209396798829 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1383.228891017911 tokens/s 1383.6172215838478 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.483099940543816 prompts/s 24.467710462959356 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3158.319892330152 tokens/s 3156.334649721757 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4918987474860657 prompts/s 0.49190931619773914 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.0285149104666 tokens/s 130.03130864371036 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.99088937654429 tokens/s 122.0263043714525 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.192169036599298 prompts/s 14.261056511482876 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3647.3874424060195 tokens/s 3665.091523451099 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9473413260984902 prompts/s 0.9476731573209913 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.32674780456017 tokens/s 279.4245893466187 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.49941426046075 tokens/s 204.5678866498336 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.983763916410926 prompts/s 9.9663761626642 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1297.8893091334203 tokens/s 1295.6289011463462 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.9664334368387 prompts/s 13.970903916114668 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3589.373393267546 tokens/s 3590.52230644147 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.47948397375122 prompts/s 6.488597325839783 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 842.3329165876586 tokens/s 843.5176523591718 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7934669887338992 prompts/s 1.800964755162785 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 233.1507085354069 tokens/s 234.12541817116204 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4637326451329037 prompts/s 0.46406416959294255 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.58308741443176 tokens/s 122.67072259019844 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.01806219683033 tokens/s 115.10028910357222 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.027995081434024 prompts/s 4.054618484357738 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4128.694958469875 tokens/s 4155.983946466682 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.037836743865684 prompts/s 12.005617992926398 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1564.918776702539 tokens/s 1560.7303390804318 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.253459713594363 prompts/s 24.18970475825256 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3128.696303053673 tokens/s 3120.47191381458 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.238139090040256 prompts/s 19.960383180376216 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2630.9580817052333 tokens/s 2594.8498134489078 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9092252611749845 prompts/s 0.9137033569406621 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.19928395274799 tokens/s 118.78143640228608 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.5997614803964915 prompts/s 5.653707334588195 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 727.968992451544 tokens/s 734.9819534964654 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0431945382581347 prompts/s 2.040568251732798 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4186.505608890918 tokens/s 4181.124347800503 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3030890123838303 prompts/s 2.303551043568131 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 711.2553027311237 tokens/s 711.3979902816674 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 461.1981809078868 tokens/s 461.2937749779968 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.5396048615443347 prompts/s 3.529499303696411 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1121.6653845747842 tokens/s 1118.4630343483557 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 809.0852389883784 tokens/s 806.8011868333552 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.451128546088046 prompts/s 2.45310324945175 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 756.9738584598844 tokens/s 757.5836995173523 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 562.7529688106571 tokens/s 563.2226936611239 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.049691648053525 prompts/s 8.054568931292382 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4129.491815451458 tokens/s 4131.9938617529915 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.111139876898282 prompts/s 8.110949351685584 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4161.014756848819 tokens/s 4160.917017414704 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.770111800073537 prompts/s 6.778380527731943 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 880.1145340095599 tokens/s 881.1894686051526 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.136682485833166 prompts/s 7.171179206337352 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3336.8272630761553 tokens/s 3352.9565497150925 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.777238240430488 prompts/s 6.777022315825469 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 881.0409712559634 tokens/s 881.012901057311 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7159286221744428 prompts/s 0.722598294088214 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.07072088267756 tokens/s 93.9377782314678 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7310781760342517 prompts/s 3.7456710255742784 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 485.04016288445274 tokens/s 486.93723332465623 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.852924058238716 prompts/s 3.8568163241392224 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1479.522838363667 tokens/s 1481.0174684694614 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9831372729780463 prompts/s 0.9838451660646144 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.88130206182024 tokens/s 290.09002669803846 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.2003489995815 tokens/s 212.35314064338635 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.881387751422338 prompts/s 28.732493074833176 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3725.699019933482 tokens/s 3706.4916066534797 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.636516049992192 prompts/s 15.572015486541934 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4018.5846248479934 tokens/s 4002.007980041277 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.235414200283952 prompts/s 3.2383631716438455 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1525.0642499310452 tokens/s 1526.4542947650727 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.342994742217844 prompts/s 7.364814804763553 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3433.290621671375 tokens/s 3443.492810115247 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2392638042240753 prompts/s 0.24076117500663205 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.104294549129794 tokens/s 31.298952750862167 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7660005784276502 prompts/s 3.7600839340665586 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3860.150592888341 tokens/s 3854.0860324182227 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.450072036437607 prompts/s 24.48509258168909 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3154.0592927004514 tokens/s 3158.5769430378928 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.983887771867096 prompts/s 0.9839046091034234 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1025891609195 tokens/s 290.10755367617475 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.3885736870438 tokens/s 212.41516605933808 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.003102702271766 prompts/s 12.006478920395466 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1560.4033512953297 tokens/s 1560.8422596514106 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.25823763611274 prompts/s 24.088141013062444 prompts/s 0.92
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1706.785446347328 tokens/s 1565.7291658490587 tokens/s 0.92
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.231767813616467 prompts/s 2.2317097461771 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 689.2294146531281 tokens/s 689.2114818793865 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 512.3186679129599 tokens/s 512.2874844284559 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.298874983122941 prompts/s 4.296001301308167 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1362.2704934018288 tokens/s 1361.359852371545 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 859.0585174607344 tokens/s 858.1577639458493 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.505296037102284 prompts/s 7.485233655042069 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3850.2168670334718 tokens/s 3839.924865036581 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7127768838214024 prompts/s 1.7130319623673949 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 558.6004854844887 tokens/s 558.6836761212692 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 389.2936323699989 tokens/s 389.3561767477934 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.774219222040995 prompts/s 5.7688721055275005 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2699.7939394574873 tokens/s 2697.2938416604384 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.881050091300708 prompts/s 11.723221360700387 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1544.5365118690918 tokens/s 1524.0187768910505 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.404162299570704 prompts/s 10.48138809173603 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1352.5410989441914 tokens/s 1362.5804519256837 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7453584394525008 prompts/s 0.7611788565110156 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.8965971288251 tokens/s 98.95325134643203 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.767640864211744 prompts/s 5.760386955245343 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2696.718162470843 tokens/s 2693.326524794513 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7691391846395232 prompts/s 3.7619252796686458 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3863.3676642555115 tokens/s 3855.9734116603618 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.474815015354076 prompts/s 0.47503736595213275 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.51260115869644 tokens/s 125.57137731578676 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.39460988078658 tokens/s 99.44115527264645 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23907648106264698 prompts/s 0.24071980892303801 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.079942538144106 tokens/s 31.29357515999494 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.909661848840227 prompts/s 0.9148041387050826 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.25604034922952 tokens/s 118.92453803166073 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.496027948651685 prompts/s 7.482753413234599 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3845.462337658314 tokens/s 3838.6525009893494 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.065397759120644 prompts/s 2.06421982745743 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4232.0000084382 tokens/s 4229.586426460274 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.818062077912565 prompts/s 26.18057918319307 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1678.1740350643165 tokens/s 1701.7376469075496 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49190112526052926 prompts/s 0.49190563133344684 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.0291434513683 tokens/s 130.03033458668335 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.95212697459041 tokens/s 121.99259657069481 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.139419661150481 prompts/s 4.164145328681783 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4242.905152679243 tokens/s 4268.248961898827 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.970460076796966 prompts/s 13.923423426897156 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3590.40823973682 tokens/s 3578.319820712569 tokens/s 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6109.549634500127 ms 6102.600778999886 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 105.80841287332987 ms 105.56953884001207 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.28574250014208 ms 66.45767050031282 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34.68305009695037 ms 34.50484903749174 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.19082203941644 ms 35.01785497536404 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13294.816539499607 ms 13438.462082000115 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 205.30474159400546 ms 207.69161295667195 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 155.3490335004426 ms 162.68896749988926 ms 0.95
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 102.75047396114238 ms 104.1323195408694 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 105.35216699315725 ms 107.03430961238783 ms 0.98
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2519.941952500176 ms 2577.4123235000843 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.46464868534167 ms 120.01311407067139 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.94256949986811 ms 83.47934300036286 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.81090724700418 ms 18.917851642837732 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.921584827928633 ms 17.005374696378382 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6044.618945000138 ms 6011.056845000212 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.79653878667311 ms 119.51665396332828 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.1993865002496 ms 76.40881049997006 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.57637550490409 ms 39.447607174651544 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.19454733194746 ms 39.04213773936851 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 250095.50855449992 ms 249174.2924559994 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 235783.59260211067 ms 235009.30902119138 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 237403.17274299992 ms 236050.11426999955 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.15392744445897 ms 68.79708311294138 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65.56227660121958 ms 65.43778767154 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1930.2846020000288 ms 1919.03614299963 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 94.30407499332432 ms 95.22951738337119 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 60.918056500213424 ms 59.753891499440215 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.252863408866046 ms 13.171030743815644 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.584415904890886 ms 12.444907479944426 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5272.785554499933 ms 5231.110777500362 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 180.98890437469285 ms 180.19965554800243 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 172.2574169998552 ms 169.80007999973168 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.08149010818899 ms 38.8138107661964 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.29161615105417 ms 36.10993344656062 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11665.552766000019 ms 11550.040795000314 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 190.46000568269238 ms 191.73528062333148 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 147.9100789997574 ms 148.74351900016336 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.80038255922813 ms 87.77569215848233 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.30587372255314 ms 89.93932269039736 ms 1.02
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57423.9524074992 ms 58115.24761200053 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24404.882942003343 ms 24693.190178945297 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 22909.76087800027 ms 22696.127297999737 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 180.74918812805373 ms 182.07596480055477 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 191.92624264264617 ms 192.91163983201352 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1977.3451549999663 ms 1942.1043275001466 ms 1.02
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 77.93893662659077 ms 80.1003962000262 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.26078049946591 ms 38.76818800017645 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.2642738624427 ms 11.053056741675453 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.368187516350444 ms 11.178593173806847 ms 1.02
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6039.057693000018 ms 6010.300804999929 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 126.99728996000356 ms 126.94023836000534 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.25761549996969 ms 86.4834220000148 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.33559771143103 ms 40.24737687956218 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.39426845817448 ms 40.325393796500656 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6372.261535500058 ms 6359.727259500005 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 108.5999934266647 ms 111.62061250000306 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.4063530000335 ms 71.44836100002294 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.344316885688436 ms 36.14834417896245 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.022484615358714 ms 36.866016742486394 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5760.337797499233 ms 5792.872616500063 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 116.97104889331482 ms 119.60485665071124 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.83178950034198 ms 88.6595430001762 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 52.470467660736524 ms 52.57606963795908 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 45.799461730583964 ms 45.77184932528473 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57082.42608699982 ms 57355.77882799987 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35345.41906373 ms 35571.797603253995 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35636.17604750004 ms 35350.50049350002 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.06400435228231 ms 96.58072383981926 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 98.20470894279043 ms 98.55851647084337 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2413.8852364999366 ms 2410.390771999573 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.33203351065336 ms 116.01512438667001 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 81.35824250030055 ms 80.54676449955878 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.909709079072172 ms 17.631069397802065 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.160069498714893 ms 16.065749860137487 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3646.939103000477 ms 3638.090063499476 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 141.82274015003713 ms 141.89584527663706 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.08160500040685 ms 92.82084300048155 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.43738117447407 ms 24.30408247226253 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.784412863099952 ms 23.61374608368908 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1841.932893500143 ms 1826.7876810000416 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.74869157995029 ms 92.30180574333039 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.3677519996163 ms 57.09178000006432 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.72703025330124 ms 12.494657948635467 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.976775719746788 ms 11.822835040136072 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7517.356016500003 ms 7477.170930999932 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 155.50804377999884 ms 155.60422404133442 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 134.28893699983746 ms 126.69303699999546 ms 1.06
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 55.86743108309888 ms 56.154255656328566 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 55.610235031058416 ms 55.70570047176915 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12615.381329000229 ms 13024.152943000445 ms 0.97
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 162.2129811033313 ms 166.09906876733658 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 129.1791285002546 ms 130.501374999767 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 112.29262829641704 ms 116.45557543230827 ms 0.96
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 101.12725211874736 ms 104.93449016074172 ms 0.96
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 73763.81392250005 ms 73612.5902235001 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 54545.00803733733 ms 54508.354054376 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66784.88195149998 ms 66708.06425700017 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.93210264944456 ms 67.84795469274688 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.24411654175546 ms 64.49598703678805 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5099.107922999792 ms 5086.0681250005655 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.4711795200104 ms 90.60802515331791 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.92975650042354 ms 63.84449399956793 ms 0.94
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.377893599611006 ms 36.191637028120056 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 30.947125218119375 ms 30.83664939242367 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2051.660143999925 ms 2027.7279659999294 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.47193586662858 ms 81.131375346716 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.246570000363135 ms 38.80655000011757 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.700552219274817 ms 11.574741119740576 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.80615936652668 ms 11.62783674047078 ms 1.02

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.608784506463351 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1249.1419858402355 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.845841491243906 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1614.9796969308538 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.02831517578501 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3186.8404864260256 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.25486336023923944 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.13223683110113 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.41500011810784 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3149.5350152359115 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.47466117127115814 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.47193401381794 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.36240518609577 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2387838181388886 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.041896358055517 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.8469884728514008 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 370.10850147068214 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.965460846275027 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1295.5099100157536 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7424705607523588 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.52117289780665 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.5945065099890305 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 857.285846298574 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.497316477435161 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 844.651142066571 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.012629280806102 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2081.6418065047933 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.080704070805357 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3106.4108251338907 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.112030953948126 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1303.0614889966214 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 820.5173979047782 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.356394742616146 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.33131654009895 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.680016366025993 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 478.40212758337907 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.928922158772549 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3579.732994804545 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.026653433823341 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4117.673211551374 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4918940760145203 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.02728005367828 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.98317226392084 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.427859332795144 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1652.8108566316844 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.31696071057386 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1471.2048923746017 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.043058481428192 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4126.089000972663 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7806290749785563 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.48177974721233 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919104348867725 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03160435796943 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.00034665771807 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.380908393074215 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3145.1371827065736 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9623382417639963 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255.1039714293195 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7937132004828635 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1456.7858689854195 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.4152251444355013 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1082.250696020166 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 780.6658244156446 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9467263559198574 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.1454217974897 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.36350694338014 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7574467992686524 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3851.3829692503687 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.783866308127265 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.90262005654444 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.021583967862203 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1691.4029579110431 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.039967978775111 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4179.894388510203 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.54344023271272 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2150.6472302526536 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.038092118065064 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3290.7303507225015 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.133602736179994 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4236.942804584494 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4477354685779167 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.9259856426894 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 561.9478452288267 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.936069451936003 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3581.5698491475528 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.451810077273244 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3971.115189859224 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.093038695930158 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1696.0475152354602 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.758008230954974 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3851.9584367288485 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23902197146510357 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.072856290463463 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.181191811467242 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3644.566295547081 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.6376668475237235 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2635.947511228192 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.076288793183771 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1309.9175431138904 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.838329850878825 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3197.3295050769034 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.776955500755706 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4054.677563694216 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.442888357833725 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 754.4290685885965 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 560.414875209442 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8110711449804755 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 575.4328422754098 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 405.51331793028834 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7135476583037863 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.76119557949222 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.0227799246117915 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4123.349422727086 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9060565163452954 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.7873471248884 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.454384298808097 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1489.0699588450527 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.623357192201408 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2629.2568887856905 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3484510482574863 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 725.2643090631987 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 538.9663843070288 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.9803865389594 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3738.469863525763 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.235871474347544 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 680.6632916651806 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.267845224970732 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1352.4374733409752 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 975.4928473811102 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.297479229947248 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1468.6722998931423 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.11817565353698 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2485.3628349598075 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.354998127500072 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.14975657500935 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.6849407988358236 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 549.522098956852 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 382.97131746126024 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3527645625465756 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.85939313105484 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9062275849985618 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.80958604981303 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.335085516304073 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1373.7452492615976 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 990.9687583999914 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.417610540661958 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3665.8717597453924 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.463311300378172 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.47170914196599 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.91355746179674 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.07080425145291 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4243.077911227012 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.781807378322924 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.63495918198012 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.90057367208776 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3572.4474337265538 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.452624288056096 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 708.8411574472924 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.1888885482800617 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1503.1336394485795 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.827086938290272 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4015.2955993429096 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0396199229456644 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4179.181222115667 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9677239110641231 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.55269005473 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.970282729953 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839662081594671 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1257163631794 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.38582614386044 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46546619583425053 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.86826556952579 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.94603091642685 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.96481427722754 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2465.42585603958 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.54803802665168 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3090.6224717323594 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23874131278739552 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.036370662361417 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0458846675910514 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4192.017683894064 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.221215384197677 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 685.970543050488 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 509.837047945149 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.29277749023713 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 708.0708297182988 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 460.52117261572266 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9053599239306973 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.69679011099065 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7561700398541262 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3850.0742908504794 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.330787539610499 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 823.002380149365 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8651209460514084 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3961.7489697026936 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9917448787428362 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4081.0852565440714 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.871022383693486 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3079.3618874964595 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839251348347453 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1136057564741 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.39663910632703 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.480067117123821 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3837.27443108452 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.482265366448731 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3838.402132988199 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.982717071386289 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.7574042218186 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.10965268801664 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0082240190487302 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 131.0691224763349 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.7263415595655647 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 354.4244027435234 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.421431270445872 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1084.2173552915924 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 782.4129029255621 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0393929717519543 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4178.7161991197545 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.635972492881956 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2635.155298771888 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.482422138329978 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 842.7148779828972 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.484648816896319 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3839.6248430678115 tokens/s

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5183.59936700017 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 89.58892007995019 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 60.16601200008154 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.652455427549434 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.203412494505784 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19158.851553500426 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1838.2070828453277 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 305.4266724993795 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 158.6336648531354 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 147.25396601399888 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1987.036208000518 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 77.9351792667027 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.638235999518656 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.397972676065994 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.461391221075985 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2063.1749590002073 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.27447359333867 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.46397750016695 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.826653004138569 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.921417902613161 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 63912.65092549975 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41293.31077156199 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40351.41946300007 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 100.20220799103322 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 102.76286338701607 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6111.512666500062 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 128.67799336666775 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 82.04266900008861 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.84021510058663 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.76792407213101 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2505.617922499823 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.91467047333815 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 81.89743850016384 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.658418541225593 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.895823970707212 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2640.063797500261 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.04558900133634 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.89289800015831 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.85509985050276 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.867666449940202 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 259946.62712449985 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 244935.409376016 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 246931.2807699996 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.38947740420745 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.10357232598422 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5439.8824734998925 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 182.17655433999607 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 173.72494299888785 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.94165538384162 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 38.111419525063 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17271.496461499737 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 222.82769583131812 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 182.16010000014649 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 142.1240172706142 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 147.67001434533415 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 77559.08805000012 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57978.830739832 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70635.4275550002 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.80080706703379 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65.31433055504868 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15648.843102000228 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.67479095332482 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 168.64950650005994 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 127.08173302458475 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 132.45753133016936 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6448.149777000026 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 111.63446464000269 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.412067999996 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.63860998310349 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.339668504378686 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6118.3996785000545 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.1081788733533 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.12022400025853 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.2819720915342 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.87209274274975 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1963.2114604996787 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.14509549332615 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57.98731200002294 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.531405476688825 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.775960485136139 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6173.967866000112 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 106.06978619333253 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.99470249982187 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.01308897676083 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.48318308686914 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7986.413046000052 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 157.93438635199405 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.75631300011992 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.89616281360583 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.539101439397896 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6058.814059999349 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.06457031861646 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.99872200062964 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 54.87317608484553 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.85837595486979 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1888.677694999842 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.74770878000406 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.50819600002433 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.058920757854036 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.31796820998434 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3702.34205450015 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 142.43163258331452 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 94.65379850007594 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.845909986718024 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.06789206535588 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70896.75558499948 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31866.693574175315 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31323.055614999248 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 193.15745061374673 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.40644514319777 ms

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7590474775400318 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3853.0236644785327 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.198002563784712 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1507.4296764809444 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.844271150712079 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 369.75524959257024 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8151179207348216 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 576.7186270336887 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 406.4230536317339 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.625123929206398 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 861.2661107968318 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.41439455300831 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2523.8712918910805 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9061873642940637 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.80435735822827 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.487086591230539 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3840.875421301266 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.448101718588022 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 756.0390934124767 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 562.2212482840507 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.029112663248922 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4118.934796246696 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.6732691824913972 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 545.715549126591 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 380.3050818477981 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.411558170588645 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3149.091004005935 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.1338287881923055 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4237.1745078971135 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.006844833977575 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4112.025064820051 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.58715614827993 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1663.1651496381955 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.88099542256637 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3112.264702466814 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9827420650945539 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.7647737000129 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.1150473300085 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7813506716139176 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.5755873098093 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4655350794998136 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.88689263114293 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.96348557805611 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.783000719010412 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.79009347135354 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9466977592844696 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.1369899842235 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.3541783191456 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0393178940146797 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4178.562364836079 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.482557167841295 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3838.551827102585 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.206673094714258 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2106.867502312854 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.41505775509987 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1651.9787540814914 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.089160212268883 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4149.739188893937 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.472561222049085 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3833.4239069111804 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0394259707685323 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4178.783814104722 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.617052474542714 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 860.2168216905527 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4438391744015515 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 754.7227060998498 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 560.6362573211111 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23899050620491202 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.068765806638563 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.219253090085445 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3654.348044151959 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.738659661802122 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.02575603427584 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.071147456813121 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4243.781139010085 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23878443695363052 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.04197680397197 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.621511984399267 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 860.7965579719047 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839210466094886 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1124003296298 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.40559580523438 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7824914893774502 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.72389361906852 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49188525562400376 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.02494847164917 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.99738109986542 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839054380093627 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.10779808185396 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.36287005514748 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.871231253630153 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4037.9416331122684 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0391518002068874 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4178.222038623912 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.679914698992896 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2558.3889108690764 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9677361737887267 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.55634651307656 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.97294341418598 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.07841078687475 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1630.0967011468585 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.348879802633779 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 825.3543743423912 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8008749671871556 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1459.5359873998677 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.953120334105916 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3585.951925865221 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.360157953096092 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.82053390249195 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 27.880746020309722 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3596.616236619954 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.934326016387006 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3581.1217862114604 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.947919962296247 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4098.615430310136 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.129767781907424 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3333.5942241086354 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.575810616987031 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4002.9833285656673 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.359650819260422 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.75460650385486 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.489444392668148 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 713.6277710468593 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.444379163391515 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1091.4893130871371 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 787.4355943123634 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.257352798410213 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 683.4558637933277 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.006536684253637 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.8497689529728 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.865356296985659 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3961.9902044103005 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.368144028537273 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1477.8587237098457 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.37677645357776 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1478.9809389651089 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.11605934909113 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1697.5438576909235 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2233236616390415 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 686.6216353451131 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 510.3239267033325 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.727863927616653 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2174.622310590165 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.64113242448022 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2637.5678763899714 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.46051683303574 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3214.9335941473228 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.08463821928587 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1311.0029685071631 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.133311569123702 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1309.8051031396099 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 824.0638386182846 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7583413426395547 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3852.299876205543 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2948393035045025 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 708.7075726369504 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 459.3533937038892 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.47468202530629644 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.4774465694664 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.36677063078473 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.653145349365781 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2643.1846395494645 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.15325627664859 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3115.770059687668 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3473917767373313 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 724.9371777705343 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 538.7232829055152 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.35419707785279 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.04562012086274 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23883535593811114 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.04859627195445 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.105952235438524 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1313.773790607008 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0620367721952912 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4225.113346228151 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.413279916356327 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3149.313109209966 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.653912440067842 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1255.0086172088195 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46333362306299347 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.47760992047169 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.91600519208363 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.946862890605078 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3248.07521313131 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.0523664788413445 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4153.675640812377 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.688238799744792 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 479.4710439668229 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.4188338568974816 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1083.394260912243 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 781.4724845644838 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9059910459021879 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.77883596728442 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.61229264810262 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3690.985751605238 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.01046683476728 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3097.3502216849793 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.899981652931261 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3572.295284803334 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.650555181539136 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2641.9735806804388 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9618783503788528 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255.04418554925087 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.2854928389817255 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1358.0298257449192 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 978.6122887480591 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.368968039063722 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1384.4822818989028 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 998.577160299123 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.512411600109154 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1496.6135080141898 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.758490735936296 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3852.4530043347036 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9054704970520556 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.71116461676722 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.7333698865066074 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 355.33808524585896 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7095124324002554 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.23661621203321 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2545207880566086 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.08770244735912 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.491911379162339 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.0318539677727 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.60705174731903 tokens/s

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 258580.5536609996 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 243620.48373112 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 245424.48084149987 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.90422552044596 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.05772969967899 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2507.259115499892 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.26669520801079 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 82.50334750027832 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.714196252413522 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.831072516258658 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80242.41102050018 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59884.95290573199 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 73712.98843349997 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.2115807755967 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.41734049939707 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3713.092553000024 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 144.17458888662318 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.94004299970038 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.834351345580902 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.104663205502337 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6160.248657500006 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 103.35623825333339 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.42573850005829 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.026456243145745 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.49456873994411 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6109.529901999963 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 127.43877080334678 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 84.11153650001779 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.91944686705478 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.81094934332291 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2630.1732105002884 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.79060928399849 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 84.27116400025625 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.820306310997374 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.78199761774397 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1965.7619674999296 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.02694897662877 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 61.73225500015178 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.561375360670501 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.853762430217106 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1994.0605810002126 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.04616736000148 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.84162300002936 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.402358041024577 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.4734086085978 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1888.522988999739 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.71816207000423 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 56.15242149997357 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.072654198259714 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.273204498764441 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6108.836415499809 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 123.25674283666103 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 82.09698599989679 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.113953054435605 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.783725132179235 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68964.95185100049 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 30565.225309190668 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29628.90754 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 191.75254985277326 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 203.0913104247209 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7910.44850849994 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 155.1658656119977 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 126.71614799990039 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.43128871719015 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.18230591285721 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17750.11726499997 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1112.1906436933145 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 227.40061249987775 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 155.6198261642097 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 143.98625150407685 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6011.335333999341 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.50289822795699 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 89.91897799933213 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 54.29048250355674 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.36026282837387 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5149.482716499733 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.22122149330002 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 61.54794949998177 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.63284431656978 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.22289830354798 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5462.653460000183 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 182.54682613734136 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.87355750047573 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41.06641913829853 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 38.31000950892242 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6491.655544000025 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 111.35712799332698 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.55617650001977 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.8136688579195 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.479692176947104 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 63265.498682500154 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40790.925173321324 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39902.949380999984 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.98898698818118 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 102.48744603708694 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17062.00229099977 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 219.39684567465397 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.65698250028072 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 139.09613905079453 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 145.31446700993249 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15098.276300499947 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 211.6495002293353 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.7528445001335 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.26872525149925 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 128.43697447887874 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2068.40576850027 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.52004253331081 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.110189999722934 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.933044359869003 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.920436219687737 ms

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.391399644496377 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1074.700633344457 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 775.2219572706562 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 27.303157883646367 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3522.1073669903813 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0684260531744316 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4238.20498295441 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9633894032697783 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255.2406224250712 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.114264322292925 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1697.4271809490401 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.914176358463065 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3575.9433241250076 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.675597365950463 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 546.4748566714414 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 380.8521106314339 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.910294863566127 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3230.9774664089787 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.345572263143476 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 724.3752634523893 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 538.3182166804804 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.228995532154451 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 679.7694191800786 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.061236835805189 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4223.4742765648325 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.090332410743216 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2481.743213396618 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.219850547926274 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 685.5490452142446 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 509.5208163651523 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9057894531076077 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.752628903989 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2918502278072204 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 707.7844663529445 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 458.6267327869585 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.38132908775959 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1649.7863907043732 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.968623824535587 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1687.9605485948132 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.466927691027308 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 840.7005998335501 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.021520520200275 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4115.040026862741 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9050651163047241 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.65846511961412 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0064508254301399 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.83860730591817 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.591308241379716 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1246.870071379363 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9826979032991886 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.75175244744344 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.10551544809687 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4633822707863846 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.49046945967291 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.93116001557769 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.345017465403104 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 434.8522705024035 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.616318665231608 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2625.965955115691 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.465406328593132 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.85207666929747 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.93086095664236 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.85951194054537 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2451.7365522708983 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.066376221023305 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3104.5625325120063 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7894279406425553 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1455.1403292067414 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.178112538873584 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3643.774922490511 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.469320661603761 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3831.7614994027294 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.389883407449293 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3146.294959560959 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.048232963089506 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4149.438787166744 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7796923023374531 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.3599993038689 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23857919542517192 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.01529540527235 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7790306292620957 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.27398180407243 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.861825251870689 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3958.370883167456 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.120697017983684 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1305.8076780288495 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 822.1587218967592 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9048595610081303 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.63174293105693 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.82754586231075 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3553.6792866138626 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7378579412425827 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.92153236153577 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.948590503127473 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1293.3167654065714 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.064853341716466 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1308.4309344231406 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.362902191599343 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3142.8143827163153 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7095957164124805 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.24744313362247 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.577482234254169 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 855.072690453042 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.348846534204494 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.3500494465842 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.233711396465214 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1460.3824815404778 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.750737648189413 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3844.506089394148 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.547456694360235 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2151.169370266831 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.304309323084958 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1469.5602120010446 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.863718672807621 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4034.087679150309 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.8446959145528146 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 369.8104688918659 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.469783921090774 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3831.999151519567 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23859420305106332 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.01724639663823 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.078919539380083 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4144.485723701982 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.918335640815934 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3577.0122596896954 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.420934504598685 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1484.7214855978293 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.616745952440694 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2626.1657375231707 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.25450395600027603 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.08551428003589 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.558035782985243 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3998.4151962272076 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0352282004593585 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4170.182582741226 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.907022309601308 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4088.1047335675357 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.447636662402331 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.8954716608372 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 562.0034858119741 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.164813862440789 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4268.934209001808 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0345774028899735 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4168.849098521556 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.426064809132061 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1085.685677365859 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 783.4770686387319 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.81597886257824 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3108.038626067586 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.603236576575554 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1664.210377477411 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.7320921465105497 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 355.17197904637146 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839754604395902 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1284444288146 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.41078264509432 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.187871219553144 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1502.6541052758873 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.083198708464983 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3311.8203881298873 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4745017802163498 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.42980058238992 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.32903932528923 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.673688460575194 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 477.5794998747752 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8069772848340286 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 574.1320973618391 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 404.5954652410945 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.442342599336855 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 707.5045379137912 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49191196389514263 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03200853604199 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.00072587218064 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.22555889733329 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3199.6613283266643 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.451983678956368 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 838.7578782643278 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9677660095093713 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.565242828806 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.97941696994667 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0348397530188147 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4169.386653935551 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.231726359122636 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1340.991765942372 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 966.8394481741851 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839928183088135 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1335624544147 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.41124971226253 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7783624124453483 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.18711361789528 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.321363628428194 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 821.7772716956651 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.616529327490755 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2626.0644523615774 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2386125579610617 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.01963253493802 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4427918794179972 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 754.399273481062 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 558.6990733812881 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.463990871703504 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3829.0273171838976 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.74834187412944 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3842.050420982676 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.02676087345815 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2970.4521526761014 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.988272961518316 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2078.475484997381 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9467696128331253 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.15817624255646 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.3728445248679 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29.14216878952025 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3759.3397738481126 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49191139048685284 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03185696129466 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.0727306632174 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.335880617253938 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1373.9972088016004 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 991.1129350143742 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7454064462533845 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3839.0416074097193 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3478594471756384 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.221728132833 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.006170571414007 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4110.6435008273 tokens/s

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66464.46923849988 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 43154.281108706666 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 42733.32001599999 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 101.34538481067432 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 103.43598965108814 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78999.34765800003 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59307.20978712267 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 72470.80709800002 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.6409276554765 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65.63519311403294 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5465.372560000105 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 183.58360737201048 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.8160190002891 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41.43363329213131 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 38.562091177908684 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8014.623977500037 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 158.14973240533376 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 131.3441125000736 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.082671469741 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.84885299357868 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6086.616079000123 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 119.41654301602344 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.9309199999625 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 54.96022618067148 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.99402132355841 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3698.734251001042 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 143.15859927336837 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.29889450034534 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.914251446910043 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.140343510939424 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6424.722442000018 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 112.49122962666434 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 71.67473500004462 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.604781314975696 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.35724412892446 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6225.111798999933 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 104.8408727333208 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.20191599990721 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.06024632703992 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.537840255846746 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18997.535682999114 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1600.7207196593154 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 278.8187685000594 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 158.32116892882993 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 145.93585365897422 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2531.0637350003162 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.37157733867086 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.57752249967598 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.655704213295667 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.870573628143323 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70453.5717635008 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31643.46789735533 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31080.083403499884 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 193.08693543780078 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 203.34199429287733 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1980.6554295000751 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.3033961066285 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 62.37369750033395 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.617845526411715 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.878749718448153 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5188.003001500874 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.74854733991137 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.3785519996527 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.8340696112454 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.335395377394576 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 260339.87888499995 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 245476.25035835197 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 246775.1600834997 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 71.45229724083546 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.49334427831721 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1989.2411519999769 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.60401550665605 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.915978000000905 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.403196774391056 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.500451686462368 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6105.555666499868 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.90624078667346 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 81.22689850006282 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.29628364690702 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.91771755614802 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18007.546417499725 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 237.73768899266724 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 184.00837800027148 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 148.99345752567072 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 153.7088325303553 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1895.7829724995463 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.05363760666235 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.2464204999269 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.119533099754277 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.366493343537526 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2645.5846254998505 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.6392997906878 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.97194100007255 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.910397253756447 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.910363340774474 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6120.664117499928 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 129.02213869667548 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.7190454999427 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.92568347358094 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.85477131355493 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2066.861660000086 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.25092127335786 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 44.316409500424925 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.824476956472038 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.916372999157963 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15749.63902599984 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.1178237659848 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 166.2659644998712 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 128.182682316079 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 133.68600323518658 ms

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7816343145010303 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.61246088513394 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.656942278306626 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2644.959931645046 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.40000271194139 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2522.0003525523807 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3595817060072126 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.7456217809376 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.432695591536753 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3966.2027670249454 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23874752395712592 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.037178114426368 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9838989496281115 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1058849610141 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.37458827722787 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7590259993671964 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3853.0016493513763 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.918377540477977 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3234.756602825883 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0398979561420965 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4179.750912135155 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9676916452607882 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.5430691450443 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.96005627223323 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.94258760324263 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3583.2450140333563 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.427510153375064 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3151.148809785383 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.246008926246041 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 681.9811604119853 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.848785399836533 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 370.34210197874927 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0400190953941455 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4179.999126462604 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0707691957779577 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4243.006082149035 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.481218046731292 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3837.864857973153 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.757576311872856 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3851.515719669677 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919051123574977 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03019740058093 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.99574723207513 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8049707629116494 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1461.1087729580734 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.020674840739835 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4121.191711758331 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.451182372355208 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 708.6537084061771 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7808192504517664 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.50650255872964 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.79462554117331 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3069.506694811357 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9827094054330354 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.7551438899486 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.10799806866638 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9043077882288956 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.56001246975643 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0396628555594716 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4179.269191041357 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.81678269119667 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4064.913151637544 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.351602871014209 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1378.9794337956928 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 994.69808746217 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839589025752542 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1235622873223 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.38752912086863 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919022540314027 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.029441830661 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.05406661863185 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.085219474246834 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1695.5392658260441 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.923161718264847 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3578.252561594066 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.356262624557446 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.31414119246796 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.937406566227764 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3732.9254470433816 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9053156385609139 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.69103301291881 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.21224967274768 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3198.7962287285995 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.25441469421849083 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.073910248403806 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.483384323690517 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3838.9761580532354 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.132249709758856 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4235.5559525028275 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.393157081511774 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1481.1104205965307 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.7358390821481335 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 355.65908067925733 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.120381244082427 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2095.6495617307155 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7417528344610682 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.42786847993887 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.483386003653423 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3838.977019874206 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.4535085512436288 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1094.3823248035935 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 789.7782659058371 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.605079117349267 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 858.6602852554047 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.438967059268009 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1089.7742714114393 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 786.0997242098371 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.448296050227745 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 756.0991082050004 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 561.5509752644366 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.085698456497163 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3107.0551008881343 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.13017000641846 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1316.9221008343998 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4744752384969024 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.42278454427117 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.32348325868489 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9056193301795076 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.730512923336 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.6833010177601304 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 478.82913230881695 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.991846868823064 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4081.294234218458 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3556689102845314 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 436.23695833698906 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.76773168040064 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2179.805118452083 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.540133561646263 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1500.2173630140142 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.648455863055647 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2640.992023330298 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8167413928310838 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 577.2344539051488 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 406.77929830323666 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.348677317094603 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 725.3341869139359 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 539.0183127034553 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0080056265507105 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 131.04073145159236 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.990863647397557 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3481.8214105142847 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.185894227647502 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3645.774816505408 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4445399621562887 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 754.939128046186 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 560.6862044933988 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.36335970860479 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1477.2367621186227 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23858217568136744 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.015682838577767 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.627074377232473 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1251.5196690402215 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4655499639171275 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.8909175760435 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.967257190174 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.045881637198438 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4192.0114746195995 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.921490893275715 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3577.823159571859 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.066717189795025 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1694.3366173366767 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.674572075728079 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 546.1404712524204 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 380.6257672874571 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.779741271402848 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.36636528237023 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9468431118045021 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.1798476592635 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.39186640153187 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.6287338463278775 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 861.735400022624 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7562523665543743 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3850.158675718234 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 48.01538489487094 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3121.000018166611 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9653106462044667 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255.49038400658068 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.1948185558558833 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1505.9288433995641 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.132518189494242 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1309.5536890688302 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 824.6522697499549 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.1127606444944105 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3325.6423669398064 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.024809372168667 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4116.727207922527 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.287572036090611 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1358.688702516754 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 979.690201577203 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2216303429243567 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 686.0986933708526 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 509.94118021969456 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.43283894348269 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1653.1345313263748 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.711686953283142 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.51930392680846 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.356208794002135 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 826.3071432202776 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4633317597709528 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.47711737785367 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.91863193679019 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.96973753643698 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3092.09614220037 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.04757358529801 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4128.405249257879 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.866276790775013 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3962.9337105443883 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.865756948753326 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2582.5484033379325 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.64384097083038 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2638.8342843214523 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2946005029870817 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 708.6338246691572 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 459.5870644769513 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.6174359185215685 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 860.2666694078039 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.118685500227198 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1697.7145575147679 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.062640585433984 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1308.143276106418 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.828546440723302 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4016.044324091054 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23881969838958653 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.04656079064625 tokens/s

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1968.3447554998565 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 94.80455332328651 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57.64442700046857 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.541103052491863 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.871362886241304 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6106.36480050016 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.76388846999937 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 77.77009599999474 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.12390406482 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.70404936507413 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2072.402735500418 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 82.25535124000089 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41.30683899984433 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.805719611744376 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.947181879927312 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3709.8409445006837 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 143.5810374200385 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.83627799979877 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.84327153864837 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.100691502900332 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15567.694932500217 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 211.95435407732356 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 170.61118799983888 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.28643066526487 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.88409230534404 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1885.0982009998916 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.19699156666744 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.964666000065336 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.972681839311665 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.240989983625319 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1985.1058935000765 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.90430167327699 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.71939650020795 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.379385445370405 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.469891548717342 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67769.83289000054 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29962.627152307323 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28717.597102499894 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 190.42693975378825 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 201.70061792719784 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 62892.62237150001 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40386.830530422 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40184.519388000124 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.63145586858826 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 101.68223204445646 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2502.4847879994923 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.04334467331257 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.55580650034244 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.610256669780377 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.822594410864983 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5157.87423349866 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.3476442533065 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 63.08617849936127 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.67354362422549 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.275710766273264 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 257234.431072 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 243333.00650803535 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 243501.93066449993 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.05580158040362 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.11879649065315 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5429.674158000125 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 183.65393957596098 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 172.55695099993318 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.96852669778389 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.996983684938925 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2651.403954000216 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 119.47014652266928 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 84.83390100036559 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.836291896778306 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.805192921584684 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6169.347012999879 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 104.59407726000487 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.40293249979368 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34.93313201739429 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.38531133968375 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79578.35537449978 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59577.81482764667 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 72953.15405799988 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.84221377872534 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.01800222368406 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6105.109807499957 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 127.2460156533369 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.1446569999298 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.946839869342064 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.72338793562835 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17937.872257999516 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1133.8330244626788 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 227.2085975000664 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 156.57282078697003 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 143.8903228539812 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16867.69092550003 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 226.9577794733262 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 177.5287084997217 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 138.74064695416712 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 144.08242617070047 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7964.563253999927 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 156.58499460933248 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 132.4896505001334 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.35169198589472 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.84237228350271 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6431.401653999956 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 110.18634970667638 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.29239400002507 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.63681207882916 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.317072432491216 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5993.155903500337 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 119.28509318669967 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.65418249974755 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 54.36501160974001 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.10.12 (main, Jun 7 2023, 13:38:31) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.33126700501665 ms

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46328670772862274 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.46520832098413 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.90745782890454 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.292132419741783 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 707.8716147474557 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 458.9429777869384 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.347374990510179 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.1587487663233 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.34193889986596 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3207.2260284912877 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.469119262472236 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3831.658181648257 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 27.7916048726088 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3585.1170285665353 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.00685451997894 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4112.0449114368475 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.954302589568968 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3586.255765519225 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.8688610862665715 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4036.725737254751 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.302053494946977 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1469.266954343107 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.14007443665441 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2488.2096767650733 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839110568257003 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1094548085805 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.39687983696393 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9615953689854666 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255.00739796811067 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.06055966015478 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3103.812196159967 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.020962432301966 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4114.753727770909 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.062613271496144 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1308.1397252944987 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.598631962664882 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1247.8221551464346 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3511090399018095 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.64417518723525 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.43665461915278 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1486.7651004898612 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.08726979678849 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1695.672536791252 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4467022191472463 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.6068906651803 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 561.7204200110759 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.362142469404144 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3142.7163785531347 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.742437440722598 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.51686729393775 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0066794459725275 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.86832797642856 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.919059804020405 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3577.198369633244 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.48697334786173 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 843.3065352220249 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.5559027611016925 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 332.26735894322 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.238826141733149 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.04739842530937 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.909507415884692 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3574.7434058823656 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7573430459511963 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3851.276622099976 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.94971504187841 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1293.462955444193 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.999761010359542 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1689.98446567337 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23886762109796603 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.052790742735585 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.315960736234649 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1471.0748957105043 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.066427612083395 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4138.077364998781 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49188653425020074 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.02528646369805 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.00097746829645 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.2501249959182905 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1346.822109956547 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 968.0679708202927 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.47190924500128 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3833.0894426856566 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9049863701473603 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.64822811915684 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.047777045827773 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4148.971471973467 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.408495276710348 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1080.1180682367421 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 779.1320289919271 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.6260333340910185 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2630.5081456875964 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.321083450107575 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 821.7408485139847 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.863044801550489 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3959.620921589251 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.47434991382851477 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.3896562214296 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.29724862810242 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.192456201732638 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1504.8153099859067 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0380230204947676 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4175.909168993779 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.790980092007751 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1455.7363553309763 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.8530711272506237 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 370.89924654258107 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.912479833115514 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4089.507317110687 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46545499789964173 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.86523749870179 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.94319343444988 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.562260503911824 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3999.5009495053387 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23882202965461627 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.046863855100117 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7805527926506222 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.47186304458089 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.058416649685157 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2087.5941644590703 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.591269937285357 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 856.8650918470966 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9677634119833676 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.5644683071606 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.9788533760712 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8038556294520143 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 573.1402517162081 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 403.90973198353794 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.88571403941515 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3112.5714125619847 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.649482036532767 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3695.783182712727 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.480253104182192 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3837.369842445464 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2202737613020327 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 685.6797447903691 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 509.6238769791782 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0381295684759664 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4176.127485807255 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7790243654516587 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.27316750871563 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.984313491760574 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3093.976440437114 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839892554433888 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.132511931668 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.39408078745547 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.468980350891289 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 710.9674456158675 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7560959135870866 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3849.998311426764 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.934260458623136 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2461.4538596210077 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3473629725991727 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 724.9282822845604 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 538.7229320288069 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7523129726193596 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3846.1207969348434 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.157243805860806 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4261.174901007326 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.483958301770942 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 842.9145792302224 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919042961451036 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.0299816429967 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.07752885531751 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.05989676350765 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4220.728468427175 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0707509801426895 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4242.968758312371 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.6605498933531404 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 541.5673140851443 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 377.44963295874226 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.447968219862037 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1654.1179342910325 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.336773890561165 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1374.2802781799276 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 990.6810628294177 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.672790473279858 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 477.46276152638154 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.632793707303523 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2633.669025786835 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.103808005343573 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1696.7475203473323 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4431076577214235 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 754.4967942419148 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 560.4716990194333 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.25454565233211796 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.09093480317534 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.351288136002162 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.667457680281 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.912733777439608 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3232.117804979663 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7093725880860688 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.21843645118895 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.233503226539855 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 680.3554194501812 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.1296870064323725 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1308.6565154683544 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 823.9936952687802 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.447853181406675 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3153.7730604014614 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.636373132067008 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2635.34262162925 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.147909003587054 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3636.012613921873 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7782767713283656 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.17598027268755 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.092340248994452 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3316.094606819846 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.650618367064023 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2164.580387718323 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9825446378738302 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.70656162589177 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.07243463868753 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9465239901463669 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.0857535746234 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.3198235929619 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.905001622237839 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.65021089091907 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.4397046180881503 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1090.007996425954 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 786.3325276498356 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.038998467839881 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4177.907860603917 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9051544911586616 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.67008385062601 tokens/s

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6440.998319999949 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 110.74126664666437 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.40421099998184 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.68560751570767 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.37614568889179 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6106.371244000002 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.95562292398245 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.63212449989078 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 55.39298069218799 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 48.46992398190285 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1885.2783004999765 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.00666148663913 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.026208000432234 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.060310247904635 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.305141051714045 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2532.953448999706 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.43352627734082 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.26540150037908 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.779799559480068 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.013203426807557 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1991.5965344998767 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 77.52473940003863 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.75204699938695 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.398765246590358 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.453427574962406 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17652.83625300026 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 224.15647835066798 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 177.60944749988994 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 146.21477302456094 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 150.62782272179493 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65234.32594349993 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 42344.713461258674 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41879.766031500025 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 100.30181319539702 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 102.8601412556169 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5174.499678500069 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.2496823665909 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 63.416820999918855 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.91975227240757 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.330177188455274 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6202.171018999934 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 105.24800641333363 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.38204749990473 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.04272655003397 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.523500103738975 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6108.172721500068 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.90046643668515 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.81508199984455 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.19418075671441 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.85883292820574 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 260871.4947394999 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 245982.4186649913 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 247559.58809399977 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.15954414937362 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.58994837999971 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7992.203209499962 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 156.57673923466547 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 134.29543700010527 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.85843149885409 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.675191881449955 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1968.6618504997568 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.32872556998578 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.17519899958279 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.543744792374257 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.818288283459308 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5493.706313000075 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 182.78887714270485 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 172.23564850064577 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41.06861954781493 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 38.19143516996267 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2072.734379499707 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 81.18704796005962 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.60844400035057 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.796053535716636 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.938045209799384 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 81030.78548899999 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 60722.77092625867 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 74389.78645499992 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.50143122540763 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.67409435002708 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15800.433527000223 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 213.21391819934192 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.30661549981596 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 127.86957135761493 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 132.48619996002597 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2631.345844000407 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 119.99958248667583 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 85.60364950017174 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.76793531848235 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.83319020247417 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18187.496590000592 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1217.637657930638 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 238.2256780001626 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 156.7914285805437 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 144.79020775133904 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3693.1756849999147 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 143.14783481670625 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.96243850005703 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.907789175550537 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.142332322012734 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6129.74736849992 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 128.53408270666893 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.85616800014395 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41.06100588782548 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.85566850307434 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69548.69400049938 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31010.123360934667 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29984.455719500147 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 191.6870091569195 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.8.17 (default, Jun 7 2023, 12:29:39) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 202.79108694443562 ms

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.905664414024263 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.7363738231542 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.050863336826483 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4152.134920247146 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.28652532960506 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1467.248292848658 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9628451385125119 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255.16986800662656 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.749248083567111 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3842.9792856562885 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.631335831782472 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2632.9873815082124 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23892154587237582 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.059800963408858 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7109605275223838 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.42486857790989 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.190922695191238 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1504.0924671415132 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.06019586251234 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2087.825462126604 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.326936905556742 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3138.1748608168195 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.459887695315286 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 709.7854003909872 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.749763029047407 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3843.507104773592 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.068802564085426 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4238.9764538110385 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.468703945827149 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3831.4451242093273 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.473503627066812 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3833.9073606852744 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0075634653303627 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.98325049294715 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.3222916646682 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1471.897916406866 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7522762527271696 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3846.0831590453486 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23877303423883933 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.040494451049113 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3465213357269987 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 724.6683623747832 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 538.536032634689 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.245348074544135 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 681.8952496907376 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.475339793688593 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 841.7941731795171 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.237204170221764 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1342.7276295015747 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 968.3056705985385 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.907041352236587 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3229.4562546517386 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.119890494967061 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1305.552098950112 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 821.4402464485125 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.593380864731594 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 857.1395124151071 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.036990654384208 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4173.793850833242 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.09008163505455 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4150.211878782985 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.108793161824791 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3323.7873307427994 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.907260186692547 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3574.1658679799843 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0064472809421594 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4111.210478650485 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.905910305179969 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4087.818948431252 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.345868453132788 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1647.481449453631 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.625241000220781 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2630.1376820632286 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.038205132028642 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4176.282315526687 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.394938693327892 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1075.8221225286757 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 776.0331928605873 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.76846789312697 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1674.950413053253 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839418731246353 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.11854109704245 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.49208691999624 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.353542729195245 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.9605547953819 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4429309183111965 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 754.4422123989857 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 559.7829623455008 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.47463516011723733 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.46505822539052 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.35696018454168 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.2210074654454126 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 685.9063321952887 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 509.79524690190374 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29.083770770521703 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3751.8064293973 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.21897673113432 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2498.4669750474613 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3508862514637077 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.615212690282 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.353101694563594 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 435.9032202932672 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.779048462351467 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.27630010569072 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49191308589196664 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03230512468247 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.99116588063512 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9827383397882414 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.76367528102895 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.11424325989404 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.603545349893113 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 338.46089548610473 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.04880973942679 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1693.1726330627414 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2547435888379577 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.11666654893451 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.868834771040428 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4036.7122375437393 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.00744534108368 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2470.9678943408785 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.54511812920504 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3682.32023866745 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.806496993262301 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 573.9794940005597 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 404.4867201047176 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.81566711839472 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3072.221058272919 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7799266337603068 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.39046238883986 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9044037701607479 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.57249012089723 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.057820686807822 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4216.474587269227 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.6775906482336167 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 547.1249404398499 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 381.31859113103206 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.079811792273718 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1310.3755329955834 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.39422730151629 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3146.8553218956013 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.331002603258348 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1372.4514149465379 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 990.2115618583003 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.59721017742782 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2157.6373230656163 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7804815735723585 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 231.4626045644066 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.677917786346333 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 478.1293122250233 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9057798331754193 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.75137831280452 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.490084003643586 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1493.7109204736662 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.15670003141707 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3638.271908074187 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8638886492003794 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3960.485865430389 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0354589114421 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4170.655309544863 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.01877535305091 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4113.631756115117 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.293124315133213 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 708.177938494873 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 459.0529128988008 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.546988129028259 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3995.5759491602626 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.850526085682443 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 370.5683911387176 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.010416138310772 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1301.3540979804004 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.795096233569694 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1457.3169536907624 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.473901937895638 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3834.111694140462 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9463796809937361 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.0432035399397 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.28236334037123 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.1652118275792835 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4269.342123268765 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4654779977769693 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.87145697219619 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.94902145003216 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.43293836493151 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1087.8638384631463 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 784.8177713604522 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7405088693123664 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.26615301060762 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23872516497317134 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.034271446512275 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.931089254003009 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3580.289938278773 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49189658816026965 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.02794411428567 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.00675041668555 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.47971033421985 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 842.3623434485804 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46334005864920075 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.47931110332974 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.9206902798991 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.00396813157704 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1625.2579285525076 tokens/s
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.633487990900844 prompts/s
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2633.993645025598 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.059169773850165 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3103.632900826671 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.610018736671345 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1249.3024357672748 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839536505404757 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.122013707361 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.4716714522085 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.888397047155955 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3112.745808065137 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.938767412658962 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3582.2632250533534 tokens/s
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.21706962145618 prompts/s
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3199.109525394652 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9677358206733926 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.5562412223233 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 209.9760925836438 tokens/s
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.330785451750097 prompts/s
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 823.0021087275127 tokens/s
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.447686441477253 prompts/s
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.9108447666151 tokens/s
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 561.9170081608939 tokens/s

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5454.024124500393 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 183.08615106667276 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.68465850045322 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 41.27397344693836 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 38.49306999706456 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17870.203808500264 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 227.24141396668105 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 185.04103700024643 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 148.66718580639292 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 154.2060986545837 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18723.386613999537 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1533.9662729066622 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 263.36673950027034 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 158.0342609946288 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 146.72360159908413 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65868.51025700013 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 42952.089562826 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 42655.900271499944 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 100.90905274563143 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 103.3437727937067 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1888.4365559997605 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.1348959600103 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.01629649952156 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.058869849482795 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.30148967918306 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2633.9105729998664 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.04579430002681 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.63644000019121 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.879160254872048 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.896625118275395 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5156.289227500565 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.94653347339529 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.3701980006881 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.684926592953 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.242309537617327 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7986.295465000012 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 158.19233767466963 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 129.38263649994042 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.88789051677662 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.71754815822603 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2075.414781500058 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.51745272664145 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.26855750023606 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.888373822442976 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.977945790411958 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3699.253162501009 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 144.4739477399647 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.98442349993275 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.714238157704624 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.115445349113315 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 259145.63174399995 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 244845.86074828464 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 246173.34830100002 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.40816401919123 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.40508523894187 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78886.52240450005 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59012.254815819986 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 72044.4028244999 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.02510526002537 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65.75853430024881 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15903.70668750029 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 213.13931954268506 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 167.40585550041942 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 128.85636857822524 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 133.88485618959274 ms
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6040.232987499621 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.90891652400994 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.51811549973354 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 54.70704003253289 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.58010815469473 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6112.463483500051 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 128.97861878998697 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.68460800003686 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.88820375273403 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.82786688691618 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6182.661464499915 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 105.88058805329638 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.79186150013084 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.09224326144527 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.473210161276306 ms
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69734.58482200021 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31249.82819040137 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 30564.476220500183 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 192.5960953923817 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 203.3323966884214 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1993.6342444998445 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.23711747331497 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.73105949944511 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.430836418906924 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.514256438506221 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6428.612557000008 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 110.42314971332341 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.48310099999344 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.62859159317332 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.34685298628225 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1975.9936359996573 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.27948016336268 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 61.18415350010764 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.555750750491628 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.861979915334244 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6117.3322049999115 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.90177221667757 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.22100400012278 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.319982478706756 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.79184074360817 ms
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2506.0447780006143 ms
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.720866174676 ms
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.5898585003888 ms
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.703319024563818 ms
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.9.17 (main, Jun 7 2023, 12:29:40) \n[GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.95419341546464 ms

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.7614254253608745 prompts/s 6.777022315825469 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 878.9853052969138 tokens/s 881.012901057311 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.354972957549521 prompts/s 5.3909289544752665 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 696.1464844814377 tokens/s 700.8207640817847 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.196706829909196 prompts/s 24.088141013062444 prompts/s 0.92
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1702.7859439440977 tokens/s 1565.7291658490587 tokens/s 0.92
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7661951050079 prompts/s 3.7619252796686458 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3860.349982633097 tokens/s 3855.9734116603618 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7959515821517156 prompts/s 1.801382309359772 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 233.47370567972303 tokens/s 234.17970021677039 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0740896386555927 prompts/s 2.0737548282729463 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4249.80966960531 tokens/s 4249.123643131267 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7647001422694055 prompts/s 3.7634225711787357 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3858.817645826141 tokens/s 3857.5081354582044 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.05753650053229 prompts/s 4.054618484357738 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4158.974913045597 tokens/s 4155.983946466682 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.769762516247443 prompts/s 5.763501051583438 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2697.710162096654 tokens/s 2694.7825516783523 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3025053475767487 prompts/s 2.303551043568131 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 711.0750514743021 tokens/s 711.3979902816674 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 460.9400805349544 tokens/s 461.2937749779968 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.029880441843858 prompts/s 12.006478920395466 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1563.8844574397015 tokens/s 1560.8422596514106 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29.254566156442838 prompts/s 28.95521254051672 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3773.839034181126 tokens/s 3735.2224177266567 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.5440724539584125 prompts/s 3.529499303696411 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1123.0811199348814 tokens/s 1118.4630343483557 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 810.120620530599 tokens/s 806.8011868333552 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.851230635250483 prompts/s 3.8568163241392224 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1478.8725639361855 tokens/s 1481.0174684694614 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9680016448981419 prompts/s 0.9682411901603802 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.6355038035783 tokens/s 288.7069306113883 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 210.02731689354985 tokens/s 210.0792910290977 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839618823433052 prompts/s 0.9839812919439876 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.12444088186464 tokens/s 290.13016386732454 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.43737039791958 tokens/s 212.41860136722823 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.624047090622705 prompts/s 10.643209396798829 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1381.1261217809517 tokens/s 1383.6172215838478 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.2340716470767275 prompts/s 3.2383631716438455 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1524.4314159959686 tokens/s 1526.4542947650727 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.571847032661611 prompts/s 5.653707334588195 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 724.3401142460095 tokens/s 734.9819534964654 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4479191808561183 prompts/s 2.4478319372827997 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.9827208931921 tokens/s 755.9557777512562 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 560.7073119979378 tokens/s 560.8048243833222 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.014067610409407 prompts/s 12.005617992926398 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1561.8287893532229 tokens/s 1560.7303390804318 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.994216162507026 prompts/s 13.92541738641096 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3596.5135537643055 tokens/s 3578.8322683076167 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9081734411139706 prompts/s 0.9137033569406621 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.06254734481617 tokens/s 118.78143640228608 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.5931751029984667 prompts/s 3.5825356745967367 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1138.641258389184 tokens/s 1135.2697299229599 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 821.6561417027967 tokens/s 819.2709765665965 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.886004150899262 prompts/s 7.844609518894351 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4045.5201294113212 tokens/s 4024.284683192802 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2392170818566456 prompts/s 0.24071980892303801 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.09822064136393 tokens/s 31.29357515999494 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.256188186651574 prompts/s 17.06325742782891 prompts/s 1.05
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2113.3044642647046 tokens/s 2218.223465617758 tokens/s 1.05
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.042554440319074 prompts/s 2.0394265863400007 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4185.1940482137825 tokens/s 4178.785075410661 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3983211707206142 prompts/s 3.407268153966697 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.78175219367984 tokens/s 442.9448600156706 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.6990004873253 prompts/s 49.425915520304486 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3230.4350316761447 tokens/s 3212.6845088197915 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.042975079920166 prompts/s 2.0402672822810786 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4186.05593875642 tokens/s 4180.50766139393 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46586631041113513 prompts/s 0.4662566113130598 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.97646188597642 tokens/s 126.08200445386888 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.0474172827789 tokens/s 118.14631692932059 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.61779604763986 prompts/s 15.572015486541934 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4013.773584243444 tokens/s 4002.007980041277 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9474087583068891 prompts/s 0.9476731573209913 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.3466304159806 tokens/s 279.4245893466187 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.61818558992755 tokens/s 204.5678866498336 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.561094470728975 prompts/s 24.48509258168909 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3168.381186724038 tokens/s 3158.5769430378928 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.007912673835737 prompts/s 2.001360152379693 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4114.213068689425 tokens/s 4100.786952225991 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7150081569948 prompts/s 0.722598294088214 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.951060409324 tokens/s 93.9377782314678 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.204504495587283 prompts/s 14.261056511482876 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3650.5576553659316 tokens/s 3665.091523451099 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.296702106333229 prompts/s 4.296001301308167 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1361.581930475937 tokens/s 1361.359852371545 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 858.5698793555767 tokens/s 858.1577639458493 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.96637140775634 prompts/s 13.923423426897156 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3589.3574517933794 tokens/s 3578.319820712569 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.844864701260924 prompts/s 28.732493074833176 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3720.9875464626593 tokens/s 3706.4916066534797 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.542748350431918 prompts/s 24.467710462959356 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3166.0145372057173 tokens/s 3156.334649721757 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.7489929711980685 prompts/s 2.796880131115109 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 357.3690862557489 tokens/s 363.59441704496413 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.846241919635131 prompts/s 1.8539691473119937 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 586.6076759536165 tokens/s 589.0628531548713 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 413.4006440211271 tokens/s 415.1123439391762 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.475809031224388 prompts/s 6.488597325839783 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 841.8551740591705 tokens/s 843.5176523591718 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839245053450746 prompts/s 0.9839046091034234 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.1134201493464 tokens/s 290.10755367617475 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.3637057369786 tokens/s 212.41516605933808 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0112377248071733 prompts/s 1.018507334458996 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 131.46090422493253 tokens/s 132.4059534796695 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.048678018228165 prompts/s 8.054568931292382 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4128.971823351048 tokens/s 4131.9938617529915 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2391604066143932 prompts/s 0.24061529216630784 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.090852859871113 tokens/s 31.27998798162002 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.23945491974350572 prompts/s 0.24076117500663205 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.12913956665574 tokens/s 31.298952750862167 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.226014556760376 prompts/s 19.960383180376216 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2629.3818923788485 tokens/s 2594.8498134489078 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.494148231912396 prompts/s 7.492620536692848 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3844.498042971059 tokens/s 3843.714335323431 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7083484155654303 prompts/s 1.7130319623673949 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 557.1561966567346 tokens/s 558.6836761212692 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 388.2870946770355 tokens/s 389.3561767477934 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.172287450616337 prompts/s 4.164145328681783 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4276.594636881746 tokens/s 4268.248961898827 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.776733089148376 prompts/s 17.724458048825532 prompts/s 1.06
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2180.975301589289 tokens/s 2304.1795463473195 tokens/s 1.06
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.693868045712232 prompts/s 26.18057918319307 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1670.101422971295 tokens/s 1701.7376469075496 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.634436481632229 prompts/s 6.7729068231492615 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 862.4767426121898 tokens/s 880.4778870094041 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.452730494453189 prompts/s 2.45310324945175 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 757.4685828336635 tokens/s 757.5836995173523 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 563.3235181220529 tokens/s 563.2226936611239 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.364860838369397 prompts/s 10.48138809173603 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1347.4319089880216 tokens/s 1362.5804519256837 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.773921337059148 prompts/s 6.778380527731943 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 880.6097738176893 tokens/s 881.1894686051526 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.49436079538075 prompts/s 21.103234859219995 prompts/s 1.03
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2664.2669033994976 tokens/s 2743.4205316985995 tokens/s 1.03
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7653237710933953 prompts/s 3.7600839340665586 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.45686537073 tokens/s 3854.0860324182227 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.998702488337036 prompts/s 13.970903916114668 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3597.6665395026184 tokens/s 3590.52230644147 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3945683874031483 prompts/s 3.409341889682664 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.2938903624093 tokens/s 443.2144456587463 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.116104510921424 prompts/s 8.110949351685584 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4163.561614102691 tokens/s 4160.917017414704 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.394322090487314 prompts/s 3.409011020079548 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.26187176335077 tokens/s 443.17143261034124 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.8866264243834343 prompts/s 2.9294087791973236 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 375.26143516984644 tokens/s 380.8231412956521 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49190012373651576 prompts/s 0.49190931619773914 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.02887870851058 tokens/s 130.03130864371036 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.9977893549724 tokens/s 122.0263043714525 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.231303023777977 prompts/s 2.2317097461771 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 689.0858751566067 tokens/s 689.2114818793865 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 512.1554456538066 tokens/s 512.2874844284559 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.7936889446848485 prompts/s 1.8035648075244035 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 233.1795628090303 tokens/s 234.46342497817247 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.851063857059662 prompts/s 11.723221360700387 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1540.638301417756 tokens/s 1524.0187768910505 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.495888767339461 prompts/s 7.485233655042069 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3845.390937645144 tokens/s 3839.924865036581 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.792890292290741 prompts/s 1.800964755162785 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 233.07573799779635 tokens/s 234.12541817116204 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.765649925394146 prompts/s 5.7688721055275005 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2695.787279117287 tokens/s 2697.2938416604384 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.87231570253984 prompts/s 3.873385234131611 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3969.1235951033364 tokens/s 3970.219864984901 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4638251583190689 prompts/s 0.46406416959294255 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.60754235006267 tokens/s 122.67072259019844 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.0410079340176 tokens/s 115.10028910357222 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.910893471626252 prompts/s 23.67526420447888 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3084.5052578397863 tokens/s 3054.109082377775 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.063705358579224 prompts/s 2.06421982745743 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4228.532279728831 tokens/s 4229.586426460274 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.392042495579219 prompts/s 7.364814804763553 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3456.22338923302 tokens/s 3443.492810115247 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9785533943080484 prompts/s 1.9917095136303464 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 257.2119412600463 tokens/s 258.92223677194505 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.545248272942333 prompts/s 4.542116862078641 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1440.3437252126957 tokens/s 1439.3514124241005 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1038.9104279119442 tokens/s 1038.1765107757747 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7489046221432982 prompts/s 0.7611788565110156 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 97.35760087862877 tokens/s 98.95325134643203 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.731258697653162 prompts/s 3.7456710255742784 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 485.06363069491107 tokens/s 486.93723332465623 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.4960619012285985 prompts/s 4.468047229100103 prompts/s 0.99
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1424.7570558803304 tokens/s 1415.8794864295314 tokens/s 0.99
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1028.2823279315894 tokens/s 1021.09772675701 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.47480323927947854 prompts/s 0.47503736595213275 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.50948827113736 tokens/s 125.57137731578676 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.3921447558375 tokens/s 99.44115527264645 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919094396839498 prompts/s 0.49190563133344684 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.0313412860553 tokens/s 130.03033458668335 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.08864353329179 tokens/s 121.99259657069481 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.494061630452785 prompts/s 7.482753413234599 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3844.453616422279 tokens/s 3838.6525009893494 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 48.10532834805965 prompts/s 47.68354528965047 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3126.8463426238773 tokens/s 3099.4304438272807 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.223051651900075 prompts/s 26.10767148263572 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1704.4983573735049 tokens/s 1696.998646371322 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.620326387587127 prompts/s 25.564807771210155 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1665.3212151931632 tokens/s 1661.71250512866 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.959126506522116 prompts/s 15.836319601059488 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4101.495512176184 tokens/s 4069.9341374722885 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.983509755835632 prompts/s 0.9838451660646144 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 289.9911298739889 tokens/s 290.09002669803846 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.2807456995628 tokens/s 212.35314064338635 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.25509024360833904 prompts/s 0.2569725363697384 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.16173166908408 tokens/s 33.40642972806599 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9099241168851357 prompts/s 0.9148041387050826 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.29013519506765 tokens/s 118.92453803166073 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.186204902956304 prompts/s 7.171179206337352 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3359.9819644262498 tokens/s 3352.9565497150925 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.760492347437578 prompts/s 5.760386955245343 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2693.3758019679144 tokens/s 2693.326524794513 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.043761102654692 prompts/s 2.040568251732798 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4187.666499339463 tokens/s 4181.124347800503 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.973032688982272 prompts/s 9.9663761626642 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1296.4942495676955 tokens/s 1295.6289011463462 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.196968551243167 prompts/s 24.18970475825256 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3121.4089431103685 tokens/s 3120.47191381458 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3571743950997774 prompts/s 2.3587430760558656 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 727.9583111906807 tokens/s 728.4427617014128 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 540.9683807762055 tokens/s 541.328390964053 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9099583239677556 prompts/s 0.9144514774798328 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.29458211580823 tokens/s 118.87869207237826 tokens/s 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: fdd9a31 Previous: fec3563 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5785.776880499725 ms 5792.872616500063 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.92931541067568 ms 119.60485665071124 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.1869515009821 ms 88.6595430001762 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 52.67676156071734 ms 52.57606963795908 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 45.66763363645848 ms 45.77184932528473 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 55960.890349500005 ms 57355.77882799987 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34505.53588642667 ms 35571.797603253995 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34309.34696250006 ms 35350.50049350002 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.76908291292905 ms 96.58072383981926 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 98.35385544190784 ms 98.55851647084337 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6048.551416500004 ms 6011.056845000212 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.46771771000446 ms 119.51665396332828 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.17563349999546 ms 76.40881049997006 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.69668735925634 ms 39.447607174651544 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.31330238613524 ms 39.04213773936851 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1927.273313000569 ms 1919.03614299963 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.58514570668073 ms 95.22951738337119 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57.797591000507964 ms 59.753891499440215 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.246371649076861 ms 13.171030743815644 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.561082803845053 ms 12.444907479944426 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2554.296359000091 ms 2577.4123235000843 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.69095240934499 ms 120.01311407067139 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.05938650028111 ms 83.47934300036286 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.02549530720225 ms 18.917851642837732 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.146301608164475 ms 17.005374696378382 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57331.99107350083 ms 58115.24761200053 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24318.87264188733 ms 24693.190178945297 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 22855.887930999415 ms 22696.127297999737 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 180.4469279026636 ms 182.07596480055477 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 191.98281539278796 ms 192.91163983201352 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6119.579037499989 ms 6102.600778999886 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 104.05635790664746 ms 105.56953884001207 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.90187099979994 ms 66.45767050031282 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34.697206180365264 ms 34.50484903749174 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.21231925674887 ms 35.01785497536404 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6024.258447000079 ms 6010.300804999929 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 126.78459783999871 ms 126.94023836000534 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 81.81675449998238 ms 86.4834220000148 ms 0.95
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.272888658516614 ms 40.24737687956218 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.24340105675399 ms 40.325393796500656 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12620.563315500476 ms 13024.152943000445 ms 0.97
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 167.80186781333032 ms 166.09906876733658 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 129.6815070008961 ms 130.501374999767 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 113.07529442774695 ms 116.45557543230827 ms 0.97
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 102.00880712124115 ms 104.93449016074172 ms 0.97
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 250191.674357 ms 249174.2924559994 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 236125.77282334934 ms 235009.30902119138 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 236989.25359399983 ms 236050.11426999955 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.64188447174384 ms 68.79708311294138 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 65.79395197603682 ms 65.43778767154 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1853.0404435000492 ms 1826.7876810000416 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 91.11207146999732 ms 92.30180574333039 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 60.19081950034888 ms 57.09178000006432 ms 1.05
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.796614877419039 ms 12.494657948635467 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.038254724045085 ms 11.822835040136072 ms 1.02
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 74282.63615700007 ms 73612.5902235001 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 55073.640529414675 ms 54508.354054376 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67166.00807750001 ms 66708.06425700017 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.38270194958277 ms 67.84795469274688 ms 0.98
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.41100075899841 ms 64.49598703678805 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2391.326898999523 ms 2410.390771999573 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 114.25296086932576 ms 116.01512438667001 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.60975599962694 ms 80.54676449955878 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.76224480426048 ms 17.631069397802065 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.083277643203157 ms 16.065749860137487 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1966.413384000134 ms 1942.1043275001466 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 77.18339140666407 ms 80.1003962000262 ms 0.96
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.758860000394634 ms 38.76818800017645 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.235902758044592 ms 11.053056741675453 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.286106823696699 ms 11.178593173806847 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7498.087665499952 ms 7477.170930999932 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 155.2915395866703 ms 155.60422404133442 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 124.13391999996293 ms 126.69303699999546 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 56.17845654502635 ms 56.154255656328566 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 55.71056960190177 ms 55.70570047176915 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6369.020522500023 ms 6359.727259500005 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 109.99698670667082 ms 111.62061250000306 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.12162149999313 ms 71.44836100002294 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.287103541900656 ms 36.14834417896245 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.95729965848386 ms 36.866016742486394 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11410.404378999829 ms 11550.040795000314 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 188.55270861600667 ms 191.73528062333148 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 146.18165550018603 ms 148.74351900016336 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 86.01591651072756 ms 87.77569215848233 ms 0.98
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.16611659868101 ms 89.93932269039736 ms 0.98
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12309.922719000042 ms 13438.462082000115 ms 0.92
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 202.7500242380041 ms 207.69161295667195 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 154.75827799991748 ms 162.68896749988926 ms 0.95
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 96.52570002857267 ms 104.1323195408694 ms 0.93
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 98.80388002495819 ms 107.03430961238783 ms 0.92
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5111.47914150024 ms 5086.0681250005655 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.57914246011205 ms 90.60802515331791 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59.33024400019349 ms 63.84449399956793 ms 0.93
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.41309699864617 ms 36.191637028120056 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 30.981834447155244 ms 30.83664939242367 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2054.349911999452 ms 2027.7279659999294 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.28242346670837 ms 81.131375346716 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.542881499983196 ms 38.80655000011757 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.664829629106839 ms 11.574741119740576 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.797083416327892 ms 11.62783674047078 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3659.3875905009554 ms 3638.090063499476 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 142.56452539669831 ms 141.89584527663706 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 89.14418849963113 ms 92.82084300048155 ms 0.96
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.47520356026676 ms 24.30408247226253 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.816849227068797 ms 23.61374608368908 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5258.380299500459 ms 5231.110777500362 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 181.02851416803605 ms 180.19965554800243 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.04110500076786 ms 169.80007999973168 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.12063777152027 ms 38.8138107661964 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.4.0", "python_version": "3.11.4 (main, Jun 7 2023, 10:57:56) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.32876665983879 ms 36.10993344656062 ms 1.01

This comment was automatically generated by workflow using github-action-benchmark.

Please sign in to comment.