use mc to fetch llm from s3

Dockerfile

@@ -1,12 +1,12 @@
-FROM inseefrlab/onyxia-vscode-pytorch:py3.12.2-gpu
+FROM inseefrlab/onyxia-python-pytorch:py3.12.3-gpu
 
 USER root
 
 WORKDIR /app
 
 COPY . /app/
 
-RUN pip install --no-cache-dir --upgrade -r requirements-app.txt && \
+RUN pip install --no-cache-dir --upgrade -r requirements.txt && \
     chmod +x entrypoint.sh
 
 EXPOSE 8000

entrypoint.sh

@@ -4,12 +4,13 @@
 export MC_HOST_s3=https://$AWS_ACCESS_KEY_ID:$AWS_SECRET_ACCESS_KEY@$AWS_S3_ENDPOINT
 mc cp --recursive s3/$S3_BUCKET/data/chroma_database/chroma_db/ data/chroma_db
 
-# Cache LLM
-python -m src.model_building.fetch_llm_model 2>&1
+# Used cached LLM from S3 if available
+MODEL_NAME_HF=$(echo "$MODEL_NAME" | sed 's|/|--|g' | sed 's|^|models--|')
+MODEL_PATH_S3=s3/$S3_BUCKET/models/hf_hub/$MODEL_NAME_HF
+if mc ls $MODEL_PATH_S3 > /dev/null 2>&1; then
+    echo "Fetching cached LLM $MODEL_NAME from S3."
+    mc cp --recursive $MODEL_PATH_S3/ $HOME/.cache/huggingface/hub/$MODEL_NAME_HF
+fi
 
 # Run app
 chainlit run app.py --host 0.0.0.0 --port 8000 -h
-
-# Run dev app 
-chainlit run dev_app.py --host 0.0.0.0 --port 8000 -h -w
-