wip(tpu): training working, tgi not working #2
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: Install docker | |
run: | | |
apt-get update -y | |
apt-get install -y docker.io | |
- name: Create test server Dockerfile | |
run: | | |
cat << EOF > Dockerfile | |
FROM python:3.9-slim | |
WORKDIR /app | |
RUN pip install flask | |
COPY server.py . | |
EXPOSE 80 | |
CMD ["python", "server.py"] | |
EOF | |
- name: Create minimal test server | |
run: | | |
cat << EOF > server.py | |
from flask import Flask, request, jsonify | |
app = Flask(__name__) | |
@app.route('/generate', methods=['POST']) | |
def generate(): | |
return jsonify({ | |
"generated_text": "Hello World!", | |
"request_received": request.json | |
}) | |
if __name__ == '__main__': | |
app.run(host='0.0.0.0', port=80) | |
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 | |
# Wait for server to start | |
sleep 5 | |
# Test the endpoint | |
curl --max-time 30 localhost:80/generate \ | |
-X POST \ | |
-d '{"inputs":"test message","parameters":{"max_new_tokens":20}}' \ | |
-H 'Content-Type: application/json' | |
# Clean up | |
docker stop test-server |