build: divide dockerfile

build/Dockerfile.cpu

@@ -0,0 +1,27 @@
+FROM python:3.10
+WORKDIR /app/
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN \
+  apt-get update && \
+  apt-get install -y software-properties-common && \
+  add-apt-repository ppa:deadsnakes/ppa && \
+  apt-get install -y python3.10 python3-pip curl && \
+  curl -sSL https://install.python-poetry.org | python3 -
+
+ENV PATH "/root/.local/bin:$PATH"
+
+COPY pyproject.toml .
+COPY poetry.lock .
+
+COPY api api
+RUN poetry env use 3.10
+RUN poetry config installer.max-workers 10
+RUN poetry install --with tools
+
+COPY . .
+
+ENV PORT 8000
+
+ENTRYPOINT ["poetry", "run", "serve"]

build/Dockerfile.gpu

@@ -17,10 +17,11 @@ COPY poetry.lock .
 
 COPY api api
 RUN poetry env use 3.10
+RUN poetry config installer.max-workers 10
 RUN poetry install --with tools
 
 COPY . .
 
-ENV PORT 8000
+ENV PORT 8001
 
 ENTRYPOINT ["poetry", "run", "serve"]

docker-compose.yml

@@ -5,14 +5,29 @@ services:
     container_name: eval
     image: eval
     build:
-      dockerfile: Dockerfile
+      dockerfile: build/Dockerfile.cpu
+      context: .
+    volumes:
+      - ./static/:/app/static/
+    ports:
+      - "3000:3000"
+      - "7000:7000"
+      - "8000:8000" # eval port
+    env_file:
+      - .env
+
+  eval.gpu:
+    container_name: eval.gpu
+    image: eval.gpu
+    build:
+      dockerfile: build/Dockerfile.gpu
       context: .
     volumes: # if you want to decrease your model download time, use this.
       - ../.cache/huggingface/:/root/.cache/huggingface/
       - ./static/:/app/static/
     ports:
-      - "7500:7500"
-      - "7501:7501"
+      - "3000:3000"
+      - "7000:7000"
       - "8000:8000" # eval port
     env_file:
       - .env
@@ -21,5 +36,5 @@ services:
         reservations:
           devices:
             - driver: nvidia
-              device_ids: ["3"] # You can choose which GPU to use
+              device_ids: ["1"] # You can choose which GPU to use
               capabilities: [gpu]