Fixed cache issue in dockerfile.

Dockerfile

@@ -10,6 +10,14 @@ RUN apt-get update && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/*
 
+# Create cache directory and set permissions
+RUN mkdir -p /app/.cache/huggingface && \
+    chmod 777 /app/.cache/huggingface
+
+# Set environment variables for cache
+ENV TRANSFORMERS_CACHE=/app/.cache/huggingface/hub
+ENV HF_HOME=/app/.cache/huggingface
+
 # Copy requirements first to leverage Docker cache
 COPY requirements.txt .
 
@@ -19,8 +27,9 @@ RUN pip install --no-cache-dir -r requirements.txt
 # Copy the rest of the application
 COPY . .
 
-# Create checkpoints directory
-RUN mkdir -p /app/checkpoints
+# Create checkpoints directory with proper permissions
+RUN mkdir -p /app/checkpoints && \
+    chmod 777 /app/checkpoints
 
 # The token will be passed during build time
 ARG HF_TOKEN
@@ -37,10 +46,14 @@ RUN if [ -n "$HF_TOKEN" ]; then \
 # Set environment variables
 ENV LLM_ENGINE_HOST=0.0.0.0
 ENV LLM_ENGINE_PORT=8001
+
+# Update MODEL_PATH for the new model
 ENV MODEL_PATH=/app/checkpoints/meta-llama/Llama-2-3b-chat-hf
 
-# Expose the port the app runs on
-EXPOSE 8001
+# Expose both ports:
+# 8001 for FastAPI
+# 7860 for Hugging Face Spaces
+EXPOSE 8001 7860
 
 # Command to run the application
 CMD ["python", "main/main.py"]