Dockerfile.gpu

# syntax=docker/dockerfile:1.7
# GPU variant of the NMT microservice (NLLB on CUDA).
# PyTorch cu118 wheels bundle CUDA 11.x libs via pip nvidia-* packages (no CUDA 12 runtime base).
FROM python:3.12-slim AS builder

WORKDIR /app

RUN apt-get update && apt-get install -y --no-install-recommends \
    gcc \
    libc6-dev \
    && rm -rf /var/lib/apt/lists/*

COPY --from=shared . /tmp/dablja-worker
RUN pip install --no-cache-dir /tmp/dablja-worker

# Install CUDA-enabled PyTorch first so device probing sees the GPU at runtime.
RUN pip install --no-cache-dir --timeout 600 --retries 3 \
    "torch==2.6.0+cu118" \
    --index-url https://download.pytorch.org/whl/cu118

COPY requirements-base.txt .
RUN pip install --default-timeout=600 --retries=3 -r requirements-base.txt

COPY --from=docker_torch verify_gpu_torch.py /tmp/docker-torch/
ENV TORCH_CUDA_VERSION=2.6.0
RUN python /tmp/docker-torch/verify_gpu_torch.py

FROM python:3.12-slim

WORKDIR /app

RUN apt-get update && apt-get install -y --no-install-recommends \
    curl \
    && rm -rf /var/lib/apt/lists/*

COPY --from=builder /usr/local /usr/local

COPY . .
RUN rm -f app/dablja_worker.py

ENV PYTHONUNBUFFERED=1 \
    PYTHONDONTWRITEBYTECODE=1

EXPOSE 8002

CMD ["python", "-m", "app.main"]