deploy.cuda.yml

proxy_location: EveryNode
http_options:
  host: 0.0.0.0
  port: 8000

applications:
  - name: video_to_audio
    route_prefix: /video_to_audio
    import_path: main:app_builder
    args:
      model: video_to_audio
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - requests==2.26.0
  - name: audio_amp_analysis
    route_prefix: /audio_amp_analysis
    import_path: main:app_builder
    args:
      model: audio_amp_analysis
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - librosa==0.10.1
  - name: audio_freq_analysis
    route_prefix: /audio_freq_analysis
    import_path: main:app_builder
    args:
      model: audio_freq_analysis
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - librosa==0.10.1
  - name: audio_rms_analysis
    route_prefix: /audio_rms_analysis
    import_path: main:app_builder
    args:
      model: audio_rms_analysis
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - librosa==0.10.1
  - name: clip_image_embedding
    route_prefix: /clip_image_embedding
    import_path: main:app_builder
    args:
      model: clip_image_embedding
      data_path: /data
      params: 
        model: xlm-roberta-base-ViT-B-32
        pretrained: laion5b_s13b_b90k
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - "open-clip-torch==2.24.0"
              - git+https://github.com/SpringsteinM/imageio.git
              - torch==2.2.0
              - scikit-learn==1.4.0
              - transformers==4.37.2
  - name: clip_text_embedding
    route_prefix: /clip_text_embedding
    import_path: main:app_builder
    args:
      model: clip_text_embedding
      data_path: /data
      params: 
        model: xlm-roberta-base-ViT-B-32
        pretrained: laion5b_s13b_b90k
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - "open-clip-torch==2.24.0"
              - git+https://github.com/SpringsteinM/imageio.git
              - torch==2.2.0
              - scikit-learn==1.4.0
              - transformers==4.37.2
  - name: clip_ontology_probs
    route_prefix: /clip_ontology_probs
    import_path: main:app_builder
    args:
      model: clip_ontology_probs
      data_path: /data
      params: 
        model: xlm-roberta-base-ViT-B-32
        pretrained: laion5b_s13b_b90k
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - "open-clip-torch==2.24.0"
              - git+https://github.com/SpringsteinM/imageio.git
              - torch==2.2.0
              - scikit-learn==1.4.0
              - transformers==4.37.2
  - name: clip_probs
    route_prefix: /clip_probs
    import_path: main:app_builder
    args:
      model: clip_probs
      data_path: /data
      params: 
        model: xlm-roberta-base-ViT-B-32
        pretrained: laion5b_s13b_b90k
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - "open-clip-torch==2.24.0"
              - git+https://github.com/SpringsteinM/imageio.git
              - torch==2.2.0
              - scikit-learn==1.4.0
              - transformers==4.37.2
  - name: color_analyser
    route_prefix: /color_analyser
    import_path: main:app_builder
    args:
      model: color_analyser
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - scikit-learn==1.4.0
  - name: color_brightness_analyser
    route_prefix: /color_brightness_analyser
    import_path: main:app_builder
    args:
      model: color_brightness_analyser
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
  - name: deepface_emotion
    route_prefix: /deepface_emotion
    import_path: main:app_builder
    args:
      model: deepface_emotion
      data_path: /data
      params: 
        model_path: /models/deepface_emotion/facial_expression_model.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - onnxruntime-gpu==1.17.0
              - onnx==1.15.0
  - name: insightface_video_detector_torch
    route_prefix: /insightface_video_detector_torch
    import_path: main:app_builder
    args:
      model: insightface_video_detector_torch
      data_path: /data
      params: 
        model_path: /models/insightface_detector_torch/scrfd_10g_bnkps.pth
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - torch==2.2.0
  - name: insightface_image_detector_torch
    route_prefix: /insightface_image_detector_torch
    import_path: main:app_builder
    args:
      model: insightface_image_detector_torch
      data_path: /data
      params: 
        model_path: /models/insightface_detector_torch/scrfd_10g_bnkps.pth
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - torch==2.2.0
  - name: insightface_facesize
    route_prefix: /insightface_facesize
    import_path: main:app_builder
    args:
      model: insightface_facesize
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_cpus: 2
          runtime_env:
            pip:
              - scikit-learn==1.4.0
  - name: insightface_video_feature_extractor
    route_prefix: /insightface_video_feature_extractor
    import_path: main:app_builder
    args:
      model: insightface_video_feature_extractor
      data_path: /data
      params: 
        model_path: /models/insightface_feature_extraction/w600k_r50.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - onnxruntime-gpu==1.17.0
              - onnx==1.15.0
  - name: insightface_image_feature_extractor
    route_prefix: /insightface_image_feature_extractor
    import_path: main:app_builder
    args:
      model: insightface_image_feature_extractor
      data_path: /data
      params: 
        model_path: /models/insightface_feature_extraction/w600k_r50.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - onnxruntime-gpu==1.17.0
              - onnx==1.15.0
  - name: min_max_norm
    route_prefix: /min_max_norm
    import_path: main:app_builder
    args:
      model: min_max_norm
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
  - name: places_classifier
    route_prefix: /places_classifier
    import_path: main:app_builder
    args:
      model: places_classifier
      data_path: /data
      params: 
        model_path: /models/places_classification/resnet50_places365.pt
        classes_file: /models/places_classification/categories_places365.txt
        hierarchy_file: /models/places_classification/scene_hierarchy_places365.csv
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - torch==2.2.0

  - name: scalar_threshold
    route_prefix: /scalar_threshold
    import_path: main:app_builder
    args:
      model: scalar_threshold
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
              
  - name: shot_annotator
    route_prefix: /shot_annotator
    import_path: main:app_builder
    args:
      model: shot_annotator
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
  - name: shot_density
    route_prefix: /shot_density
    import_path: main:app_builder
    args:
      model: shot_density
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip: 
              - scikit-learn==1.4.0
  - name: shot_scalar_annotator
    route_prefix: /shot_scalar_annotator
    import_path: main:app_builder
    args:
      model: shot_scalar_annotator
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
  - name: shot_type_classifier
    route_prefix: /shot_type_classifier
    import_path: main:app_builder
    args:
      model: shot_type_classifier
      data_path: /data
      params: 
        model_path: /models/shot_type_classification/shot_type_classifier_e9-s3199.pt
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - torch==2.2.0
  - name: thumbnail_generator
    route_prefix: /thumbnail_generator
    import_path: main:app_builder
    args:
      model: thumbnail_generator
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20

            
  - name: transnet_shotdetection
    route_prefix: /transnet_shotdetection
    import_path: main:app_builder
    args:
      model: transnet_shotdetection
      data_path: /data
      params: 
        model_path: /models/transnet_shotdetection/transnet.pt
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - torch==2.2.0

  - name: video_to_video
    route_prefix: /video_to_video
    import_path: main:app_builder
    args:
      model: video_to_video
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20


  - name: aggregate_list_scalar_per_time
    route_prefix: /aggregate_list_scalar_per_time
    import_path: main:app_builder
    args:
      model: aggregate_list_scalar_per_time
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20

  - name: aggregate_scalar
    route_prefix: /aggregate_scalar
    import_path: main:app_builder
    args:
      model: aggregate_scalar
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20

  - name: cosine_similarity
    route_prefix: /cosine_similarity
    import_path: main:app_builder
    args:
      model: cosine_similarity
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          runtime_env:
            pip:
              - scipy==1.12.0
  
  - name: aggregate_scalar_per_time
    route_prefix: /aggregate_scalar_per_time
    import_path: main:app_builder
    args:
      model: aggregate_scalar_per_time
      data_path: /data
      params: {}
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
  - name: whisper
    route_prefix: /whisper
    import_path: main:app_builder
    args:
      model: whisper
      data_path: /data
      params: 
        model: openai/whisper-medium
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - librosa==0.10.1
              - torch==2.2.0
              - scikit-learn==1.4.0
              - transformers==4.37.2


  - name: x_clip_video_embedding
    route_prefix: /x_clip_video_embedding
    import_path: main:app_builder
    args:
      model: x_clip_video_embedding
      data_path: /data
      params: 
        model_path: /models/xclip/xclip_16_8_video.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - onnxruntime-gpu==1.17.0
              - onnx==1.15.0
              - cuda-python==12.2.0
              - regex==2023.12.25
              - ftfy==6.1.3
  - name: x_clip_text_embedding
    route_prefix: /x_clip_text_embedding
    import_path: main:app_builder
    args:
      model: x_clip_text_embedding
      data_path: /data
      params: 
        model_path: /models/xclip/xclip_16_8_text.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - onnxruntime-gpu==1.17.0
              - onnx==1.15.0
              - cuda-python==12.2.0
              - regex==2023.12.25
              - ftfy==6.1.3
  - name: x_clip_probs
    route_prefix: /x_clip_probs
    import_path: main:app_builder
    args:
      model: x_clip_probs
      data_path: /data
      params: 
        text_model_path: /models/xclip/xclip_16_8_text.onnx
        sim_model_path: /models/xclip/xclip_16_8_sim.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scikit-learn==1.4.0
              - opencv-python-headless==4.9.0.80
              - onnxruntime-gpu==1.17.0
              - onnx==1.15.0
              - cuda-python==12.2.0
              - regex==2023.12.25
              - ftfy==6.1.3
  - name: face_clustering
    route_prefix: /face_clustering
    import_path: main:app_builder
    args:
      model: face_clustering
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scipy==1.12.0  
  - name: place_clustering
    route_prefix: /place_clustering
    import_path: main:app_builder
    args:
      model: place_clustering
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_gpus: 0.1
          runtime_env:
            pip:
              - scipy==1.12.0  

  - name: clustering
    route_prefix: /clustering
    import_path: main:app_builder
    args:
      model: clustering
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_cpus: 2
          runtime_env:
            pip:
              - scipy==1.12.0  
  - name: cluster_size_filter
    route_prefix: /cluster_size_filter
    import_path: main:app_builder
    args:
      model: cluster_size_filter
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_cpus: 2
           
  - name: face_size_filter
    route_prefix: /face_size_filter
    import_path: main:app_builder
    args:
      model: face_size_filter
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          num_cpus: 2
                             
              
  - name: blip_image_embedding
    route_prefix: /blip_image_embedding
    import_path: main:app_builder
    args:
      model: blip_image_embedding
      data_path: /data
      params: 
        model: Salesforce/instructblip-flan-t5-xl
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          # num_cpus: 2
          num_gpus: 0.1
          runtime_env:
            pip:
              - git+https://github.com/SpringsteinM/imageio.git
              - torch==2.2.0
              - pillow
              - scikit-learn==1.4.0
              - transformers==4.37.2   
              - accelerate      
              - bitsandbytes
              - peft   
  - name: blip_vqa
    route_prefix: /blip_vqa
    import_path: main:app_builder
    args:
      model: blip_vqa
      data_path: /data
      params: 
        model: Salesforce/instructblip-flan-t5-xl
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
          max_replicas: 20
        ray_actor_options:
          # num_cpus: 2
          num_gpus: 1.0
          runtime_env:
            pip:
              - git+https://github.com/SpringsteinM/imageio.git
              - torch==2.2.0
              - pillow
              - scikit-learn==1.4.0
              - transformers==4.37.2   
              - accelerate      
              - bitsandbytes
              - peft   
  - name: timeline_video_sampler
    route_prefix: /timeline_video_sampler
    import_path: main:app_builder
    args:
      model: timeline_video_sampler
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
        ray_actor_options:
          num_cpus: 1
  - name: invert_scalar
    route_prefix: /invert_scalar
    import_path: main:app_builder
    args:
      model: invert_scalar
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
        ray_actor_options:
          num_cpus: 1

  - name: dbscanclustering
    route_prefix: /dbscanclustering
    import_path: main:app_builder
    args:
      model: dbscanclustering
      data_path: /data
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 0
        ray_actor_options:
          num_cpus: 1
          runtime_env:
            pip:
              - scipy==1.12.0
              - scikit-learn==1.4.0
  - name: ocr_video_detector_onnx
    route_prefix: /ocr_video_detector_onnx
    import_path: main:app_builder
    runtime_env:
      pip:
        - "--extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/"
        - scikit-learn
        - opencv-python-headless
        - onnxruntime-gpu==1.17.0
        - onnx
        - torch
        - torchvision
        - pytorch_lightning
        - timm
        - nltk
    args:
      model: ocr_video_detector_onnx
      data_path: /data
      params: 
        model_path: /models/ocr/yolov7_text_det_dyn_input.onnx
    deployments:
      - name: Deployment
        autoscaling_config:
          min_replicas: 1
        ray_actor_options:
          num_gpus: 0.2