Dockerfile

# Use an official Python image as a base
FROM python:3.10

# Set the working directory
WORKDIR /app

# Copy requirements and install dependencies
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

ENV TRANSFORMERS_CACHE=/app/cache

# Copy the application code
COPY . .

# Download the Hugging Face model during the build process
# RUN python -c "from transformers import AutoTokenizer, AutoModelForCausalLM; \
#     model_path = 'model/models--meta-llama--Llama-3.2-3B-Instruct/snapshots/0cb88a4f764b7a12671c53f0838cd831a0843b95'; \
#     AutoTokenizer.from_pretrained(model_path); \
#     AutoModelForCausalLM.from_pretrained(model_path)"

# Expose the default port
EXPOSE 7860

# Start the FastAPI app using Uvicorn
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]