William Valentin
636f4b3311
Add Dockerfile and K8s manifests for whisper.cpp transcription service. Deploys to ai-stack namespace with worker node affinity for GPU access.
41 lines
1.2 KiB
Docker
41 lines
1.2 KiB
Docker
FROM ubuntu:22.04 AS build
|
|
WORKDIR /app
|
|
|
|
RUN apt-get update && \
|
|
apt-get install -y build-essential wget cmake git \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
ARG WHISPER_CPP_VERSION=v1.8.3
|
|
RUN git clone --depth 1 --branch ${WHISPER_CPP_VERSION} \
|
|
https://github.com/ggml-org/whisper.cpp.git .
|
|
|
|
ARG WHISPER_MODEL=base.en
|
|
RUN bash ./models/download-ggml-model.sh ${WHISPER_MODEL}
|
|
RUN cmake -B build && cmake --build build --config Release -j$(nproc)
|
|
|
|
FROM ubuntu:22.04
|
|
WORKDIR /app
|
|
|
|
RUN apt-get update && \
|
|
apt-get install -y curl ffmpeg \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
COPY --from=build /app/build/bin/whisper-server /usr/local/bin/whisper-server
|
|
COPY --from=build /app/build/src/libwhisper.so* /usr/local/lib/
|
|
COPY --from=build /app/build/ggml/src/libggml*.so* /usr/local/lib/
|
|
COPY --from=build /app/models /app/models
|
|
RUN ldconfig
|
|
|
|
ARG WHISPER_MODEL=base.en
|
|
ENV WHISPER_MODEL_PATH=/app/models/ggml-${WHISPER_MODEL}.bin
|
|
|
|
EXPOSE 8080
|
|
|
|
ENTRYPOINT ["whisper-server"]
|
|
CMD ["--model", "/app/models/ggml-base.en.bin", \
|
|
"--host", "0.0.0.0", \
|
|
"--port", "8080", \
|
|
"--convert", \
|
|
"--language", "en", \
|
|
"--inference-path", "/v1/audio/transcriptions"]
|