Dockerfile

FROM python:3.10-slim

# 1) Install CUDA runtime 11.7
RUN apt-get update && apt-get install -y gnupg curl && \
    curl -s -L https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.0-1_all.deb \
      -o cuda-keyring.deb && \
    dpkg -i cuda-keyring.deb && \
    apt-get update && \
    apt-get install -y cuda-runtime-11-7 && \
    rm -rf /var/lib/apt/lists/* cuda-keyring.deb

# install build/c runtime deps
RUN apt-get update && apt-get install -y \
      build-essential python3-dev git \
    && rm -rf /var/lib/apt/lists/*

WORKDIR /app

# create a cache dir for pip / torch / pooch / numba
ENV \
  XDG_CACHE_HOME=/app/.cache \
  PIP_CACHE_DIR=/app/.cache/pip \
  TORCH_HOME=/app/.cache/torch \
  HOME=/app \
  NUMBA_CACHE_DIR=/app/.cache/numba \
  NUMBA_DISABLE_CACHING=1

RUN mkdir -p /app/.cache/{pip,torch,numba} \
 && chmod -R 777 /app/.cache

# 2) install ALL your Python deps (including CPU-only torch)
COPY requirements.txt .
RUN pip install --upgrade pip \
 && pip install --no-cache-dir -r requirements.txt

# 2) install torchmcubes (from Git or local wheel)
#    Option A: from Git
# RUN pip install --no-cache-dir \
#       git+https://github.com/tatsy/torchmcubes.git#egg=torchmcubes

#    Option B: from a wheel directory in your repo
COPY wheel /wheel
RUN pip install --no-cache-dir /wheel/torchmcubes-*.whl

# copy and expose
COPY . .
EXPOSE 8000

CMD ["uvicorn","src.main:app","--host","0.0.0.0","--port","8000"]