scott/rocm-chatterbox-whisper
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
47 Commits 1 Branch 0 Tags
e0eb7c36df0699b690ba93b95e01cc72bb89c920
Commit Graph

3 Commits

Author SHA1 Message Date
scott
fe3c77ff4f Upgrade to ROCm 7.2, Python 3.11, PyTorch 2.11.0 
- Base image: rocm/dev-ubuntu-22.04:6.1 → 7.2
- Python 3.10 → 3.11 via deadsnakes PPA
- torch/torchaudio: 2.5.1 → 2.11.0
- torchvision: 0.20.1 → 0.26.0
- pytorch_triton_rocm: 3.1.0 → 3.3.0
- transformers: 4.46.3 → >=4.50.0
- diffusers: 0.29.0 → >=0.32.0
- safetensors: >=0.4.1 → >=0.4.5
- config: temperature 0.8→0.9, seed 0→1960

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-04-06 17:09:56 -04:00
scott
59731084cd Multi-pass warmup and smaller chunk_size to fix HA timeout
All checks were successful
Build ROCm Image / build (push) Successful in 2m49s
Details 
torch.compile with dynamic=True still specializes per shape family on
first call. The warmup was running one text length, leaving real requests
to JIT-compile their own shapes (15-22s for first chunk). HA freezes
because it gets no AudioChunk for 22 seconds.

Fix:
- Run 3 warmup passes (short/medium/long text) so torch.compile builds
  a dynamic shape graph covering the range HA actually sends. Real
  requests then hit a cached compilation and synthesize in 3-8s.
- Reduce default chunk_size from 300 to 120 chars so the first text
  chunk is shorter, producing faster synthesis and earlier first audio.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-04-05 15:04:46 -04:00
scott
16ea2853f5 Initial implementation: Chatterbox TTS with ROCm and Wyoming
All checks were successful
Build ROCm Image / build (push) Successful in 15m27s
Details 
Wyoming-only server built around the official chatterbox TTS model.
Includes ROCm/AMD GPU support, sentence-level streaming, config.yaml
management, and Gitea CI for container builds.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-04-05 09:51:09 -04:00