3 people 0d31687638 fix(tada): run voice-prompt encode under torch.inference_mode (#955) ...

Encoder.eval() alone still builds an autograd graph because parameters
require grad by default. On 8GB GPUs that ballooned TADA encode VRAM far
past the model footprint (issue 890). Wrap the encode forward in
inference_mode and add a unit test that asserts the flag is set.

Co-authored-by: fooSynaptic <[email protected]>

2026-07-26 23:32:25 -07:00

..

__init__.py

fix(backend): close model load races

2026-07-26 23:17:20 -07:00

base.py

chore(backend): repair test suite and bring ruff to green

2026-07-26 23:16:09 -07:00

chatterbox_backend.py

chore(backend): repair test suite and bring ruff to green

2026-07-26 23:16:09 -07:00

chatterbox_turbo_backend.py

chore(backend): repair test suite and bring ruff to green

2026-07-26 23:16:09 -07:00

hume_backend.py

fix(tada): run voice-prompt encode under torch.inference_mode (#955)

2026-07-26 23:32:25 -07:00

kokoro_backend.py

fix(backend): close model load races

2026-07-26 23:17:20 -07:00

luxtts_backend.py

fix(backend): close model load races

2026-07-26 23:17:20 -07:00

mlx_backend.py

fix(mlx): fail generation when voice cloning fails

2026-07-26 23:17:08 -07:00

pytorch_backend.py

chore(backend): repair test suite and bring ruff to green

2026-07-26 23:16:09 -07:00

qwen_custom_voice_backend.py

chore(backend): repair test suite and bring ruff to green

2026-07-26 23:16:09 -07:00

qwen_llm_backend.py

chore(backend): repair test suite and bring ruff to green

2026-07-26 23:16:09 -07:00