Use gunicorn as WSGI server in ML image

machine-learning/Dockerfile

@@ -6,7 +6,7 @@ ENV PYTHONDONTWRITEBYTECODE=1 \
 
 RUN python -m venv /opt/venv && \
     /opt/venv/bin/pip install --pre torch  -f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html && \
-    /opt/venv/bin/pip install transformers tqdm numpy scikit-learn scipy nltk sentencepiece flask Pillow && \
+    /opt/venv/bin/pip install transformers tqdm numpy scikit-learn scipy nltk sentencepiece flask Pillow gunicorn && \
     /opt/venv/bin/pip install --no-deps sentence-transformers
 
 FROM python:3.10-slim
@@ -22,4 +22,4 @@ WORKDIR /usr/src/app
 
 COPY . .
 
-CMD ["python", "src/main.py"]
+CMD ["gunicorn", "src.main:server"]

machine-learning/gunicorn.conf.py

@@ -0,0 +1,13 @@
+"""
+Gunicorn configuration options.
+https://docs.gunicorn.org/en/stable/settings.html#config-file
+"""
+import os
+
+
+# Set the bind address based on the env
+server_port = os.getenv('MACHINE_LEARNING_PORT') or "3003"
+bind = f"127.0.0.1:{server_port}"
+
+# Preload the Flask app / models etc. before starting the server
+preload_app = True