Nik Afiq
99b624ef20
All checks were successful
CI / test (push) Successful in 6s
CI / build-ai-gateway (push) Successful in 50s
CI / build-ha-gateway (push) Successful in 50s
CI / build-discord-bot (push) Successful in 1m3s
CI / build-tts-gateway (push) Successful in 5m32s
CI / build-tts-sidecar (push) Successful in 17m17s
31 lines
1.2 KiB
Docker
31 lines
1.2 KiB
Docker
# Inference sidecar for tts-gateway (TTS_GATEWAY_PLAN.md Phase 3). Base image
|
|
# matches nik-gpu's driver 570.211.01 / CUDA 12.8 - only meaningfully runnable
|
|
# on a CUDA host, even though it builds fine anywhere. Does NOT bake in model
|
|
# weights: CHECKPOINT_PATH/CONFIG_PATH must point at a mounted G_790000.pth +
|
|
# uma.json at runtime (model artifact distribution is still an open decision,
|
|
# TTS_GATEWAY_PLAN.md Phase 5 - for now this means a bind mount, see the repo
|
|
# README/CLAUDE.md for the exact nik-gpu run command).
|
|
FROM pytorch/pytorch:2.9.1-cuda12.8-cudnn9-runtime
|
|
|
|
RUN apt-get update && apt-get install -y --no-install-recommends \
|
|
build-essential \
|
|
cmake \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
WORKDIR /app
|
|
|
|
COPY requirements.txt .
|
|
RUN pip install --no-cache-dir -r requirements.txt
|
|
|
|
COPY . .
|
|
|
|
# models/models.py does a module-level `import monotonic_align`, even though
|
|
# this sidecar's infer()-only path never calls it (only the training-time
|
|
# forward() does) - the extension still has to build for the import itself to
|
|
# succeed. See tmp/reference/uma-tts-api/spike/FINDINGS.md for the identical
|
|
# issue hit during the Phase 0 spike.
|
|
RUN cd monotonic_align && python setup.py build_ext --inplace
|
|
|
|
EXPOSE 50054
|
|
CMD ["python", "server.py"]
|