wip(tpu): training working, tgi not working #7
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Debug DIND | |
| on: | |
| push: | |
| jobs: | |
| debug-dind: | |
| runs-on: | |
| group: gcp-ct5lp-hightpu-8t | |
| container: | |
| image: us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla:r2.4.0_3.10_tpuvm | |
| options: --shm-size "16gb" --ipc host --privileged ${{ vars.V5_LITEPOD_8_ENV}} -v /mnt/hf_cache:/mnt/hf_cache -e PJRT_DEVICE=TPU | |
| steps: | |
| - name: Checkout code | |
| uses: actions/checkout@v4 | |
| - name: Install docker | |
| run: | | |
| apt-get update -y | |
| apt-get install -y docker.io | |
| # the run test step is similar to "Create test server Dockerfile", "Create minimal test server", "Build and run test container" | |
| # but it's not in a single step inside a python script so it can be run locally | |
| # this scripts work on the local machine, not in the github actions environment | |
| # i left commented out for now, as "Create test server Dockerfile", "Create minimal test server", "Build and run test container" maybe be easier to debug in a ci environment | |
| # - name: Run test | |
| # run: | | |
| # python debug-dind-locally/test.py | |
| - name: Create test server Dockerfile | |
| run: | | |
| cat << EOF > Dockerfile | |
| FROM us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla:r2.4.0_3.10_tpuvm | |
| WORKDIR /app | |
| RUN pip install fastapi uvicorn | |
| COPY server.py . | |
| EXPOSE 80 | |
| CMD ["uvicorn", "server:app", "--host", "0.0.0.0", "--port", "80"] | |
| EOF | |
| - name: Create minimal test server | |
| run: | | |
| cat << EOF > server.py | |
| from fastapi import FastAPI | |
| from pydantic import BaseModel | |
| app = FastAPI() | |
| class GenerateRequest(BaseModel): | |
| inputs: str | |
| @app.post("/generate") | |
| async def generate(request: GenerateRequest): | |
| return { | |
| "generated_text": "Hello World!", | |
| "request_received": request.dict() | |
| } | |
| EOF | |
| - name: Build and run test container | |
| run: | | |
| docker build -t test-tgi-server . | |
| docker run -d -p 80:80 --name test-server test-tgi-server | |
| sleep 5 | |
| docker logs -f test-server & | |
| sleep 10 | |
| # Test the endpoint | |
| curl --max-time 30 localhost:80/generate \ | |
| -X POST \ | |
| -d '{"inputs":"test message"}' \ | |
| -H 'Content-Type: application/json' | |
| # Clean up | |
| docker stop test-server |