| FROM docker.io/nvidia/cuda:12.3.2-cudnn9-devel-ubuntu22.04@sha256:fb1ad20f2552f5b3aafb2c9c478ed57da95e2bb027d15218d7a55b3a0e4b4413 | |
| WORKDIR /app | |
| COPY requirements.txt . | |
| # Install torch FIRST, before flash-attn | |
| RUN pip install --no-cache-dir torch | |
| # VERIFY torch installation - Add this line | |
| RUN python -c "import torch; print(torch.__version__)" | |
| # Now install all other requirements, including flash-attn | |
| RUN pip install --no-cache-dir -r requirements.txt | |
| COPY . . | |
| CMD ["python", "app.py"] |