evals-moss / Dockerfile
bezzam's picture
bezzam HF Staff
Update Dockerfile
0270649 verified
Raw
History Blame Contribute Delete
1.62 kB
FROM nvidia/cuda:12.9.0-runtime-ubuntu24.04
# Avoid interactive prompts during package installation
ENV DEBIAN_FRONTEND=noninteractive
# Install Python and system dependencies. ffmpeg/libsndfile1 for audio I/O (torchcodec/librosa).
# git is handy for any source deps.
RUN apt-get update && apt-get install -y --no-install-recommends \
python3 \
python3-pip \
python3-dev \
git \
libsndfile1 \
ffmpeg \
&& rm -rf /var/lib/apt/lists/*
# Set Python alias (Ubuntu 24.04 ships Python 3.12)
RUN ln -sf /usr/bin/python3 /usr/bin/python
ENV PIP_BREAK_SYSTEM_PACKAGES=1
WORKDIR /app
# Install PyTorch (cu128 wheels). MOSS-TTS's [torch-runtime] extra pins torch 2.9.1 + torchcodec.
RUN pip install --no-cache-dir \
torch==2.9.1 \
torchaudio==2.9.1 \
torchcodec==0.8.1 \
--index-url https://download.pytorch.org/whl/cu128
# Runtime deps for the MOSS-TTS remote code (loaded via trust_remote_code). transformers MUST
# be 5.0.0 (the remote processor/generate target the 5.x API). The model + its modeling/processing
# code are downloaded from the HF hub at load time into the mounted cache.
RUN pip install --no-cache-dir \
"transformers==5.0.0" \
"accelerate>=1.10.1" \
safetensors==0.6.2 \
numpy==2.1.0 \
orjson==3.11.4 \
einops==0.8.1 \
scipy==1.16.2 \
librosa==0.11.0 \
tiktoken==0.12.0 \
soundfile \
datasets \
tqdm
# Copy the full repository
COPY . /app
# Default entrypoint
ENTRYPOINT ["bash"]
# Keep-alive CMD so the Space runtime stays healthy; `docker run` overrides it.
EXPOSE 7860
CMD ["-c", "python3 -m http.server 7860"]