flow-jcloud.yml

jtype: Flow
with:
  protocol: grpc
  port: 51005
  monitoring: true  # enable prometheus & grafana
  env:
    JINA_LOG_LEVEL: debug
  exit_on_exceptions:
    - RuntimeError
    - ValueError
jcloud:
  gateway:
    ingress: kong
    timeout: 6000
executors:
  - name: dalle
    uses: jinahub+docker://DalleGenerator/latest
    timeout_ready: -1  # slow download speed often leads to timeout
    env:
      XLA_PYTHON_CLIENT_ALLOCATOR: platform  # https://jax.readthedocs.io/en/latest/gpu_memory_allocation.html
    replicas: 1  # change this if you have larger VRAM
    gpus: all
    jcloud:
      resources:
        memory: 10G
        gpu: shared
  - name: clip_encoder
    # uses: jinahub+docker://CLIPTorchEncoder/latest-gpu
    uses: jinahub+docker://CLIPOnnxEncoder/latest-gpu
    uses_with:
      name: ViT-L-14-336::openai
    replicas: 1
    needs: [gateway]
    gpus: all
    jcloud:
      resources:
        memory: 5G
        gpu: shared
  - name: diffusion
    uses: jinahub+docker://GLID3Diffusion/latest
    timeout_ready: -1  # slow download speed often leads to timeout
    replicas: 1  # change this if you have larger VRAM
    needs: [clip_encoder]
    gpus: all
    jcloud:
      resources:
        memory: 10G
        gpu: shared
  - name: rerank
    # uses: jinahub+docker://CLIPTorchEncoder/latest-gpu
    uses: jinahub+docker://CLIPOnnxEncoder/latest-gpu
    uses_with:
      name: ViT-L-14-336::openai
    uses_requests:
      '/': rank
    replicas: 1
    needs: [dalle, diffusion]
    gpus: all
    jcloud:
      resources:
        memory: 5G
        gpu: shared
  - name: upscaler
    uses: jinahub+docker://SwinIRUpscaler/latest
    replicas: 1
    gpus: all
    jcloud:
      resources:
        memory: 5G
        gpu: shared
  - name: store
    uses: jinahub+docker://DalleFlowStore/latest
    floating: true
    replicas: 1
    jcloud:
      resources:
        memory: 500M