scott dc179ad8c6 Fix input dtype mismatches; drop MIGraphXExecutionProvider ...

- Drop MIGraphXExecutionProvider — symbol mismatch with apt migraphx,
  ROCMExecutionProvider handles GPU execution fine without it
- Add _ort_type_to_np() helper to read expected dtypes from session metadata
- prepare_voice: cast audio to session's declared input dtype (float32, not float16)
- _generate: read embed_dtype and kv_dtype from LM session metadata

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

2026-04-06 20:11:57 -04:00

.gitea/workflows

Fix CI: add main trigger, tag onxx builds as :onxx not :latest

2026-04-06 19:10:02 -04:00

.gitignore

Initial implementation: Chatterbox TTS with ROCm and Wyoming

2026-04-05 09:51:09 -04:00

config.py

Multi-pass warmup and smaller chunk_size to fix HA timeout

2026-04-05 15:04:46 -04:00

config.yaml

Multi-pass warmup and smaller chunk_size to fix HA timeout

2026-04-05 15:04:46 -04:00

docker-compose.yml

Update docker-compose.yml

2026-04-06 19:14:43 -04:00

Dockerfile.rocm

Replace individual ROCm libs with rocm-libs meta-package; graceful provider fallback

2026-04-06 19:34:28 -04:00

engine.py

Fix input dtype mismatches; drop MIGraphXExecutionProvider

2026-04-06 20:11:57 -04:00

main.py

Cache voice conditionals and add FP16 autocast

2026-04-05 20:22:13 -04:00

perth_stub.py

Replace resemble-perth with a no-op stub

2026-04-05 11:16:08 -04:00

requirements-rocm-init.txt

Switch back to PyPI onnxruntime-rocm — AMD wheel too old for ONNX IR v10

2026-04-06 19:52:33 -04:00

requirements-rocm.txt

Add MIGraphX/half deps and use AMD onnxruntime wheel for ROCm 6.1.3

2026-04-06 19:23:28 -04:00

wyoming_handler.py

Switch to ONNX runtime with chatterbox-turbo-ONNX (fp16)

2026-04-06 19:08:34 -04:00

wyoming_voices.py

Add supports_synthesize_streaming=True to TtsProgram

2026-04-05 16:52:02 -04:00

Description

No description provided

213 KiB

Languages

Python 100%