Merge pull request #2 from pharmbio/feature/scilife_serve_bundle

Scilifelab serve version in single container..

Author: morganekmefjord

README.md

@@ -46,13 +46,34 @@ To use this project, follow these steps:
     docker-compose up --build
     ```
 
+### Scilifelab Deployment
+
+1. Build container
+    ```sh
+    docker build -t ptp .
+    ```
+2. Push container to registry
+
+
+3. Deploy
+
+4. Configuration
+    - Set environment variables
+    Use `DOWNLOAD=true` to download models prior to startup.
+ - preferebly mount `/app/inference/models` directory to outside storage to avoid ram bloat.
+
+ - Use `MAX_MODELS` to limit the number of active models (for debug purposes) .otherwise it iterates all models present (currently 800+).
+
+- If desired use `MODEL_DIR` to change path where model are found (default /app/inference/models/models)
+
 
 ## References
 Model repository on Hugging Face: [pharmbio/ptp](https://huggingface.co/pharmbio/ptp)
 
 ## License
 This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
 
+
 ## Authors
 - **Jonathan Alvarsson** - *Research, scipipe version and models* - [jonathanalvarsson](https://github.com/jonalv) 
 - **Morgan Ekmefjord** - *Web service, deployment and service packaging* - [morganekmefjord](https://github.com/morganekmefjord)

ptp/.dockerignore

@@ -1,4 +1,5 @@
 */conformal_models/*
 */vennABERS_models/*
 inference/models/*
-*.jar
+*.jar
+tmp/*

ptp/Dockerfile

@@ -1,12 +1,14 @@
 FROM python:3.11 AS djangobase
 
-# Set working directory
+ENV USER=ptp
 
 RUN mkdir /app
+
 WORKDIR /app
 
 # Install system dependencies
-RUN apt-get update && apt-get install -y gcc python3-dev libpq-dev
+RUN apt-get update && apt-get install -y gcc git-lfs python3-dev redis-server supervisor \
+    && apt-get clean
 
 # Install Java (OpenJDK)
 RUN apt-get update && apt-get install -y openjdk-17-jre-headless \
@@ -20,6 +22,8 @@ ENV PATH=$JAVA_HOME/bin:$PATH
 COPY requirements.txt /app/
 RUN pip install -r requirements.txt
 
+COPY supervisord.conf /etc/supervisor/conf.d/supervisord.conf
+
 # Copy the application code
 FROM djangobase AS djangoapp
 
@@ -32,5 +36,21 @@ WORKDIR /app
 # Expose the port the Django app runs on
 EXPOSE 8000
 
-# Default command
-CMD ["gunicorn", "p2p.wsgi:application", "--bind", "0.0.0.0:8000"]
+RUN useradd -u 1000 $USER
+
+RUN chown -R $USER:$USER /etc/supervisor/conf.d/supervisord.conf
+RUN chown -R $USER:$USER /app/
+RUN chown -R $USER:$USER /var/log/supervisor/
+RUN chmod +x /app/start-single.sh
+RUN chmod +x /app/start-django.sh
+RUN chmod +x /app/download.sh
+
+# Make sure the container is running as non-root
+USER $USER
+
+ENV CELERY_BROKER_URL="redis://localhost:6379/0"
+ENV REDIS_URL="redis://localhost:6379/0"
+ENV DOWNLOAD=false
+
+# Start supervisord
+CMD ["sh", "-c", "/app/start-single.sh"]

ptp/download.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+if [ "$DOWNLOAD" = "true" ]; then
+  git clone https://huggingface.co/pharmbio/ptp /app/inference/models && \
+  cd /app/inference/models && \
+  git lfs pull
+else
+  echo "noop"
+fi
+exit 0

ptp/inference/tasks.py

@@ -7,6 +7,9 @@
 import pandas as pd
 from datetime import timezone
 
+model_dir = os.environ.get("MODEL_DIR", "/app/inference/models/models/")
+max_models = os.environ.get("MAX_MODELS", False)
+
 @shared_task
 def run_inference(job_id):
 
@@ -37,7 +40,7 @@ def run_inference(job_id):
         chembl_version = 'chembl_34'
     else:
         chembl_version = 'chembl_34'
-    base = '/app/inference/models/'
+    base = model_dir
     try:
         model_files = [os.path.join(base,chembl_version,path, f) for f in os.listdir(os.path.join(base, chembl_version,path)) if f.endswith('.jar')]
     except Exception as e:
@@ -54,6 +57,8 @@ def run_inference(job_id):
     if DEBUG:
         print("RUNNING IN DEBUG MODE ONLY CALCULATING 3 MODELS", flush=True)
         model_files = model_files#[:5]
+    if max_models:
+        model_files = model_files[:int(max_models)]
 
     try:
         for model in model_files: