Upload Dockerfile

Dockerfile

@@ -1,17 +1,20 @@
 # Use the official vLLM Docker image as the base image
 FROM vllm/vllm-openai:latest
 
+# Use the root user to ensure write permissions (if needed)
+USER root
+
 # Set the working directory
 WORKDIR /app
 
 # Install system dependencies if needed (e.g., wget)
 RUN apt-get update && apt-get install -y wget && rm -rf /var/lib/apt/lists/*
 
-# (Optional) Set up writable cache directories if your app needs them
+# Create and set permissions for cache directories (for Hugging Face, matplotlib, etc.)
 RUN mkdir -p /app/.cache && chmod -R 777 /app/.cache \
     && mkdir -p /app/.config/matplotlib && chmod -R 777 /app/.config/matplotlib
 
-# Set environment variables for caching and user agent (adjust as needed)
+# Set environment variables for cache directories and user agent
 ENV HF_HOME=/app/.cache
 ENV XDG_CACHE_HOME=/app/.cache
 ENV MPLCONFIGDIR=/app/.config/matplotlib
@@ -30,14 +33,8 @@ EXPOSE 7860
 # Override the base image's entrypoint so our CMD is executed directly.
 ENTRYPOINT []
 
+# Create a writable log file (alternatively, you could redirect logs to /tmp)
+RUN touch /app/vllm.log && chmod 666 /app/vllm.log
+
 # Set the CMD to launch the vLLM server (for your new model) in the background and then start the Gradio app.
-CMD ["bash", "-c", "vllm.entrypoints.openai.api_server \
---model unsloth/llama-3-8b-Instruct-bnb-4bit \
---enable-auto-tool-choice \
---tool-call-parser llama3_json \
---chat-template examples/tool_chat_template_llama3.1_json.jinja \
---quantization bitsandbytes \
---load-format bitsandbytes \
---dtype half \
---max-model-len 8192 \
-> vllm.log 2>&1 & python app.py"]
+CMD ["bash", "-c", "vllm.entrypoints.openai.api_server --model unsloth/llama-3-8b-Instruct-bnb-4bit --enable-auto-tool-choice --tool-call-parser llama3_json --chat-template examples/tool_chat_template_llama3.1_json.jinja --quantization bitsandbytes --load-format bitsandbytes --dtype half --max-model-len 8192 > /app/vllm.log 2>&1 & python3 app.py"]