NathanParas26/LocalAI

1ARG GO_VERSION=1.21-bullseye2 3FROM golang:$GO_VERSION as requirements4 5ARG BUILD_TYPE6ARG CUDA_MAJOR_VERSION=117ARG CUDA_MINOR_VERSION=78ARG SPDLOG_VERSION="1.11.0"9ARG PIPER_PHONEMIZE_VERSION='1.0.0'10ARG TARGETARCH11ARG TARGETVARIANT12 13ENV BUILD_TYPE=${BUILD_TYPE}14ENV EXTERNAL_GRPC_BACKENDS="huggingface-embeddings:/build/extra/grpc/huggingface/huggingface.py,autogptq:/build/extra/grpc/autogptq/autogptq.py,bark:/build/extra/grpc/bark/ttsbark.py,diffusers:/build/extra/grpc/diffusers/backend_diffusers.py,exllama:/build/extra/grpc/exllama/exllama.py,vall-e-x:/build/extra/grpc/vall-e-x/ttsvalle.py"15ENV GALLERIES='[{"name":"model-gallery", "url":"github:go-skynet/model-gallery/index.yaml"}, {"url": "github:go-skynet/model-gallery/huggingface.yaml","name":"huggingface"}]'16ARG GO_TAGS="stablediffusion tts"17 18RUN apt-get update && \19    apt-get install -y ca-certificates cmake curl patch pip20 21# Use the variables in subsequent instructions22RUN echo "Target Architecture: $TARGETARCH"23RUN echo "Target Variant: $TARGETVARIANT"24 25# CuBLAS requirements26RUN if [ "${BUILD_TYPE}" = "cublas" ]; then \27    apt-get install -y software-properties-common && \28    apt-add-repository contrib && \29    curl -O https://developer.download.nvidia.com/compute/cuda/repos/debian11/x86_64/cuda-keyring_1.0-1_all.deb && \30    dpkg -i cuda-keyring_1.0-1_all.deb && \31    rm -f cuda-keyring_1.0-1_all.deb && \32    apt-get update && \33    apt-get install -y cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \34    ; fi35ENV PATH /usr/local/cuda/bin:${PATH}36 37# Extras requirements38COPY extra/requirements.txt /build/extra/requirements.txt39ENV PATH="/root/.cargo/bin:${PATH}"40RUN pip install --upgrade pip41RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y42RUN if [ "${TARGETARCH}" = "amd64" ]; then \43        pip install git+https://github.com/suno-ai/bark.git diffusers invisible_watermark transformers accelerate safetensors;\44    fi45RUN if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH}" = "amd64" ]; then \46        pip install torch && pip install auto-gptq https://github.com/jllllll/exllama/releases/download/0.0.10/exllama-0.0.10+cu${CUDA_MAJOR_VERSION}${CUDA_MINOR_VERSION}-cp39-cp39-linux_x86_64.whl;\47    fi48RUN pip install -r /build/extra/requirements.txt && rm -rf /build/extra/requirements.txt49 50# Vall-e-X51RUN git clone https://github.com/Plachtaa/VALL-E-X.git /usr/lib/vall-e-x && cd /usr/lib/vall-e-x && pip install -r requirements.txt52 53WORKDIR /build54 55# OpenBLAS requirements56RUN apt-get install -y libopenblas-dev57 58# Stable Diffusion requirements59RUN apt-get install -y libopencv-dev && \60    ln -s /usr/include/opencv4/opencv2 /usr/include/opencv261 62 63# piper requirements64# Use pre-compiled Piper phonemization library (includes onnxruntime)65#RUN if echo "${GO_TAGS}" | grep -q "tts"; then \66RUN test -n "$TARGETARCH" \67    || (echo 'warn: missing $TARGETARCH, either set this `ARG` manually, or run using `docker buildkit`')68 69RUN curl -L "https://github.com/gabime/spdlog/archive/refs/tags/v${SPDLOG_VERSION}.tar.gz" | \70    tar -xzvf - && \71    mkdir -p "spdlog-${SPDLOG_VERSION}/build" && \72    cd "spdlog-${SPDLOG_VERSION}/build" && \73    cmake ..  && \74    make -j8 && \75    cmake --install . --prefix /usr && mkdir -p "lib/Linux-$(uname -m)" && \76    cd /build && \77    mkdir -p "lib/Linux-$(uname -m)/piper_phonemize" && \78    curl -L "https://github.com/rhasspy/piper-phonemize/releases/download/v${PIPER_PHONEMIZE_VERSION}/libpiper_phonemize-${TARGETARCH:-$(go env GOARCH)}${TARGETVARIANT}.tar.gz" | \79    tar -C "lib/Linux-$(uname -m)/piper_phonemize" -xzvf - && ls -liah /build/lib/Linux-$(uname -m)/piper_phonemize/ && \80    cp -rfv /build/lib/Linux-$(uname -m)/piper_phonemize/lib/. /usr/lib/ && \81    ln -s /usr/lib/libpiper_phonemize.so /usr/lib/libpiper_phonemize.so.1 && \82    cp -rfv /build/lib/Linux-$(uname -m)/piper_phonemize/include/. /usr/include/83# \84#    ; fi85 86###################################87###################################88 89FROM requirements as builder90 91ARG GO_TAGS="stablediffusion tts"92 93ENV GO_TAGS=${GO_TAGS}94ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility95ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"96ENV NVIDIA_VISIBLE_DEVICES=all97 98WORKDIR /build99 100COPY Makefile .101RUN make get-sources102COPY go.mod .103RUN make prepare104COPY . .105COPY .git .106 107RUN ESPEAK_DATA=/build/lib/Linux-$(uname -m)/piper_phonemize/lib/espeak-ng-data make build108 109###################################110###################################111 112FROM requirements113 114ARG FFMPEG115ARG BUILD_TYPE116ARG TARGETARCH117 118ENV BUILD_TYPE=${BUILD_TYPE}119ENV REBUILD=false120ENV HEALTHCHECK_ENDPOINT=http://localhost:8080/readyz121 122# Add FFmpeg123RUN if [ "${FFMPEG}" = "true" ]; then \124    apt-get install -y ffmpeg \125    ; fi126 127WORKDIR /build128 129# we start fresh & re-copy all assets because `make build` does not clean up nicely after itself130# so when `entrypoint.sh` runs `make build` again (which it does by default), the build would fail131# see https://github.com/go-skynet/LocalAI/pull/658#discussion_r1241971626 and132# https://github.com/go-skynet/LocalAI/pull/434133COPY . .134RUN make prepare-sources135COPY --from=builder /build/local-ai ./136 137# Copy VALLE-X as it's not a real "lib"138RUN cp -rfv /usr/lib/vall-e-x/* extra/grpc/vall-e-x/139 140# To resolve exllama import error141RUN if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH:-$(go env GOARCH)}" = "amd64" ]; then \142        cp -rfv /usr/local/lib/python3.9/dist-packages/exllama extra/grpc/exllama/;\143    fi144# Define the health check command145HEALTHCHECK --interval=1m --timeout=10m --retries=10 \146  CMD curl -f $HEALTHCHECK_ENDPOINT || exit 1147 148EXPOSE 8080149ENTRYPOINT [ "/build/entrypoint.sh" ]150