35 lines
1.0 KiB
Docker
35 lines
1.0 KiB
Docker
# BiRefNet background removal service — CUDA 12.6 inference image.
|
|
# torch's +cu126 wheels bundle their own CUDA/cuDNN, so no nvidia/cuda base is
|
|
# needed; the GPU driver is injected by the nvidia container runtime at run time.
|
|
FROM python:3.12-slim
|
|
|
|
ENV PYTHONUNBUFFERED=1 \
|
|
UV_PROJECT_ENVIRONMENT=/app/.venv \
|
|
UV_PYTHON_PREFERENCE=only-system \
|
|
UV_COMPILE_BYTECODE=1 \
|
|
UV_LINK_MODE=copy \
|
|
HF_HOME=/app/hf_cache \
|
|
PORT=8000
|
|
|
|
# uv: fast, reproducible Python + dependency management.
|
|
COPY --from=ghcr.io/astral-sh/uv:0.9 /uv /uvx /bin/
|
|
|
|
WORKDIR /app
|
|
|
|
# Install dependencies first so this layer is cached across code changes.
|
|
# The BuildKit cache mount keeps the uv download cache warm across rebuilds.
|
|
COPY pyproject.toml uv.lock ./
|
|
RUN --mount=type=cache,target=/root/.cache/uv \
|
|
uv sync --no-install-project --no-dev --frozen
|
|
|
|
# Application code.
|
|
COPY src ./src
|
|
COPY README.md ./
|
|
RUN --mount=type=cache,target=/root/.cache/uv \
|
|
uv sync --no-dev
|
|
|
|
ENV PATH="/app/.venv/bin:${PATH}"
|
|
|
|
EXPOSE 8000
|
|
CMD ["rmbg-as-a-service"]
|