Dockerfile.cuda

FROM nvidia/cuda:12.8.1-devel-ubuntu24.04

ENV DEBIAN_FRONTEND=noninteractive

RUN apt-get update && apt-get install -y --no-install-recommends \
        python3.12 \
        python3.12-venv \
        ffmpeg \
        libsndfile1 \
        sox \
    && rm -rf /var/lib/apt/lists/*

COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/

WORKDIR /app

ENV UV_PYTHON=python3.12

COPY pyproject.toml uv.lock ./

# Install all dependencies (CPU torch from lock file).
RUN uv sync --frozen --no-dev

# Replace CPU torch with CUDA 12.8 variant.
ENV VIRTUAL_ENV=/app/.venv \
    PATH="/app/.venv/bin:$PATH"
RUN uv pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu128

# Install flash attention for reduced GPU memory usage.
RUN uv pip install flash-attn --no-build-isolation

COPY main.py .

ENV QWEN_TTS_DEVICE=cuda:0 \
    QWEN_TTS_DTYPE=bfloat16 \
    QWEN_TTS_ATTN=flash_attention_2 \
    HOST=0.0.0.0 \
    PORT=8000 \
    NVIDIA_VISIBLE_DEVICES=all \
    NVIDIA_DRIVER_CAPABILITIES=compute,utility \
    PYTHONWARNINGS="ignore::SyntaxWarning"

EXPOSE 8000

CMD ["/app/.venv/bin/python", "main.py"]