# ==================== BASE CONTAINER + REQUIREMENTS ==================== # NB: Make sure the CUDA version supports the driver version: # https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html FROM nvidia/cuda:11.8.0-devel-ubuntu22.04 # See: https://github.com/opencontainers/image-spec/blob/master/annotations.md LABEL org.opencontainers.image.title='Storyteller ML : VITS' LABEL org.opencontainers.image.description='Storyteller ML : VITS' LABEL org.opencontainers.image.documentation='https://github.com/storytold/storyteller-ml' LABEL org.opencontainers.image.source='https://github.com/storytold/storyteller-ml' LABEL org.opencontainers.image.url='https://github.com/storytold/storyteller-ml' # tzdata install hangs without presetting the time zone ENV TZ=UTC RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone # Notes on packages: # * espeak-ng is critical for speech # * ffmpeg: file length detection # * libmagic: mimetype detection, other magic bytes operations in Python # * screen: not sure why included; I threw in tmux and curl for good measure # * unzip: not sure why included, but perhaps downloading models RUN DEBIAN_FRONTEND=noninteractive apt-get update \ && apt-get install -y --no-install-recommends \ build-essential \ curl \ espeak-ng \ ffmpeg \ libmagic-dev \ libmagic1 \ python3-dev \ python3-pip \ python3.10 \ python3.10-venv \ screen \ tmux \ unzip RUN python3 --version # ==================== INSTALL MORE VITS REQUIREMENTS ==================== WORKDIR / RUN mkdir -p /vits COPY ./requirements.txt /vits WORKDIR /vits RUN pip3 install --upgrade pip setuptools \ && pip3 install -r requirements.txt \ && pip3 install \ transformers \ g2p-en \ scipy \ text_unidecode \ emoji \ phonemizer \ numpy \ matplotlib \ librosa==0.9.2 \ Cython \ Unidecode \ tensorboard \ gdown # ==================== VITS CODE + FINAL BUILD ==================== COPY ./ ./ RUN cd monotonic_align \ && mkdir monotonic_align \ && python3 setup.py build_ext --inplace \ && cd .. WORKDIR /