Get rid of server_args in internal iterfaces

sgl-project · Aug 17, 2024 · a65ceec · a65ceec
1 parent 4086209
commit a65ceec
Show file tree

Hide file tree

Showing 23 changed files with 615 additions and 1,598 deletions.
diff --git a/examples/usage/llm_engine.py b/examples/usage/llm_engine.py
@@ -16,5 +16,5 @@
 outputs = llm.generate(prompts, sampling_params)
 # Print the outputs.
 for prompt, output in zip(prompts, outputs):
-    print('===============================')
-    print(f"Prompt: {prompt}\nGenerated text: {output['text']}")
+    print("===============================")
+    print(f"Prompt: {prompt}\nGenerated text: {output['text']}")
diff --git a/python/sglang/__init__.py b/python/sglang/__init__.py
@@ -2,8 +2,8 @@
 
 from sglang.api import (
     LLM,
-    SamplingParams,
     Runtime,
+    SamplingParams,
     assistant,
     assistant_begin,
     assistant_end,

diff --git a/python/sglang/api.py b/python/sglang/api.py
@@ -18,8 +18,9 @@
     SglSelect,
     SglVideo,
 )
-from sglang.srt.serving.engine import LLM
 from sglang.srt.sampling_params import SamplingParams
+from sglang.srt.serving.engine import LLM
+
 
 def function(
     func: Optional[Callable] = None, num_api_spec_tokens: Optional[int] = None

diff --git a/python/sglang/bench_latency.py b/python/sglang/bench_latency.py
@@ -49,13 +49,13 @@
 import torch
 import torch.distributed as dist
 
+from sglang.srt.config import ModelConfig
 from sglang.srt.hf_transformers_utils import get_tokenizer
 from sglang.srt.managers.schedule_batch import Req, ScheduleBatch
-from sglang.srt.model_config import ModelConfig
 from sglang.srt.model_executor.forward_batch_info import ForwardMode
 from sglang.srt.model_executor.model_runner import ModelRunner
 from sglang.srt.sampling_params import SamplingParams
-from sglang.srt.server_args import ServerArgs
+from sglang.srt.serving.server_args import ServerArgs
 from sglang.srt.utils import suppress_other_loggers
 
 
@@ -111,7 +111,7 @@ def load_model(server_args, tp_rank):
     suppress_other_loggers()
     rank_print = print if tp_rank == 0 else lambda *args, **kwargs: None
 
-    model_config = ModelConfig(path=server_args.model_path)
+    model_config = ModelConfig(model_path=server_args.model_path)
     model_runner = ModelRunner(
         model_config=model_config,
         mem_fraction_static=server_args.mem_fraction_static,

diff --git a/python/sglang/launch_server.py b/python/sglang/launch_server.py
@@ -2,8 +2,8 @@
 
 import argparse
 
-from sglang.srt.server import launch_server
-from sglang.srt.server_args import ServerArgs
+from sglang.srt.serving.server import launch_server
+from sglang.srt.serving.server_args import ServerArgs
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()