Commit Graph
177 Commits
Author SHA1 Message Date
Hawk Lee ce9c5a8bf4 fix: export AIIA Audio Info node correctly (v1.4.84) 2025-12-28 23:06:58 +08:00
Hawk Lee 8a26498f67 feat: add Audio Info node for metadata inspection (v1.4.83) 2025-12-28 23:03:11 +08:00
Hawk Lee ca9353b7af docs: add high_pass_hz documentation (v1.4.82) 2025-12-28 22:58:22 +08:00
Hawk Lee 2d91854dcc fix: indentation error and variable name in high pass filter logic (v1.4.82) 2025-12-28 22:51:33 +08:00
Hawk Lee 10eeb407f5 feat: add high_pass_hz parameter to remove low-freq artifacts (v1.4.81) 2025-12-28 22:50:50 +08:00
Hawk Lee a0c8a7fa32 fix: restore missing use_cuda parameter in INPUT_TYPES (v1.4.80) 2025-12-28 22:44:09 +08:00
Hawk Lee ff1b91bfd9 docs: update README with Resemble Enhance usage guide (v1.4.79) 2025-12-28 22:42:41 +08:00
Hawk Lee 543e41fd03 feat: expose denoise_strength parameter (v1.4.79) 2025-12-28 22:31:53 +08:00
Hawk Lee 3fa7309f89 fix: lowercase solver method name for compatibility (v1.4.78) 2025-12-28 22:22:51 +08:00
Hawk Lee 1810337983 fix: use official configurate_ method for model setup (v1.4.77) 2025-12-28 22:08:05 +08:00
Hawk Lee aa73a4596b fix: remove duplicate code lines in aiia_audio_enhance.py (v1.4.76) 2025-12-28 21:52:32 +08:00
Hawk Lee 133f0d6749 fix: pad last chunk to speed up compilation (v1.4.76) 2025-12-28 21:51:47 +08:00
Hawk Lee 7544c90b41 fix: update safe_merge_chunks to support arbitrary args/kwargs (v1.4.75) 2025-12-28 21:44:44 +08:00
Hawk Lee a990e447e7 fix: re-add normalization step to define abs_max (v1.4.74) 2025-12-28 21:37:35 +08:00
Hawk Lee 0d7dc8f1c7 fix: update safe_inference_chunk signature to accept sr argument (v1.4.73) 2025-12-28 21:30:07 +08:00
Hawk Lee 20b5081774 fix: restore mangled safe_inference_chunk structure logic (v1.4.72) 2025-12-28 21:23:35 +08:00
Hawk Lee 03b090d437 fix: correct indentation in aiia_audio_enhance.py (v1.4.72) 2025-12-28 21:22:40 +08:00
Hawk Lee ddd0a9b317 fix: force update internal latent vars to prevent inference skip (v1.4.72) 2025-12-28 21:21:49 +08:00
Hawk Lee c9ddfde293 fix: promote forward pass to primary path and silence warning (v1.4.71) 2025-12-28 20:53:39 +08:00
Hawk Lee d8cb101035 feat: use Hz frequency labels for spectrogram y-axis (v1.4.70) 2025-12-28 20:49:32 +08:00
Hawk Lee 90f1378a11 fix: tensor mismatch size error and attempt to find ode_solve in lcfm submodule (v1.4.69) 2025-12-28 20:46:01 +08:00
Hawk Lee 5fd5bf1501 fix: restore ode_solve logic and add execution trace to debug inconsistency (v1.4.68) 2025-12-28 20:38:17 +08:00
Hawk Lee b83f77a75b fix: use object.__setattr__ to bypass frozen dataclass check for quality settings (v1.4.67) 2025-12-28 20:29:05 +08:00
Hawk Lee dd8410a500 fix: cleanup debugs and force inject nfe parameters to improve quality (v1.4.66) 2025-12-28 20:21:47 +08:00
Hawk Lee eb2a112f6f fix: add deep debug prints for merge_chunks device mismatch (v1.4.65) 2025-12-28 20:08:50 +08:00
Hawk Lee 78a255ef54 fix: patch merge_chunks to force cpu inputs (v1.4.64) 2025-12-28 20:07:57 +08:00
Hawk Lee 63967f550a fix: force mel_fn to cpu to reset persistent state (v1.4.63) 2025-12-28 19:59:33 +08:00
Hawk Lee cc30d4f9ce fix: revert mel_fn move to keep merge logic on CPU (v1.4.62) 2025-12-28 19:53:12 +08:00
Hawk Lee 77471807f8 fix: return cpu tensor from inference_chunk to fix merge_chunks crash (v1.4.61) 2025-12-28 19:44:35 +08:00
Hawk Lee d1b0e93d4b fix: implement correct monkey patch based on source inspection (v1.4.60) 2025-12-28 19:36:29 +08:00
Hawk Lee 0743b611f0 fix: inspect original source to debug missing ode_solve (v1.4.59) 2025-12-28 19:30:41 +08:00
Hawk Lee 32c2c236eb fix: debug model structure and force mel_fn move (v1.4.58) 2025-12-28 19:23:00 +08:00
Hawk Lee ff746e342a fix: restore missing inference call that caused UnboundLocalError (v1.4.57) 2025-12-28 18:42:32 +08:00
Hawk Lee bede92a711 fix: move global mel_fn to device to prevent STFT error (v1.4.56) 2025-12-28 18:33:38 +08:00
Hawk Lee 5bb4d64567 fix: force inject model config and robust attribute access (v1.4.55) 2025-12-28 18:16:40 +08:00
Hawk Lee d4c57282f0 fix: pass Path object to load_enhancer to avoid type error (v1.4.54) 2025-12-28 18:06:43 +08:00
Hawk Lee 1478fb8f0a fix: monkey patch inference for cuda safety and locate site-packages model (v1.4.53) 2025-12-28 18:02:25 +08:00
Hawk Lee 675fa91927 fix: active download patch and cpu fallback for resemble enhance (v1.4.52) 2025-12-28 17:35:18 +08:00
Hawk Lee 3c01f6bf2a fix: skip repetitive model download and attempt to resolve device handling (v1.4.51) 2025-12-28 17:02:40 +08:00
Hawk Lee 149789b715 fix: correctly configure resemble enhance model parameters manually (v1.4.50) 2025-12-28 16:54:41 +08:00
Hawk Lee 8f8ca6847e fix: manual model loading and device placement for Resemble Enhance (v1.4.49) 2025-12-28 16:48:46 +08:00
Hawk Lee d93a7c13ff fix: ensure 1D tensor input for resemble-enhance (v1.4.48) 2025-12-28 16:35:04 +08:00
Hawk Lee 3b5525dac2 fix: ensure deepspeed mock is treated as a package with __path__ attribute (v1.4.47) 2025-12-28 16:29:12 +08:00
Hawk Lee 5e792d2606 fix: remove shadowed sys import in _mock_deepspeed (v1.4.46) 2025-12-28 16:25:28 +08:00
Hawk Lee 58e18da734 fix: resolve solver case mismatch and optimize deepspeed mockup timing (v1.4.45) 2025-12-28 16:15:41 +08:00
Hawk Lee 98c461bb46 fix: robust DeepSpeed mocking via MetaPathFinder (v1.4.44) 2025-12-28 16:05:36 +08:00
Hawk Lee 58fec77c9b fix: robust deepspeed mock and enable fp16 for CosyVoice (v1.4.43) 2025-12-28 16:00:01 +08:00
Hawk Lee 4024380af3 fix: explicit mock for deepspeed.accelerator to fix import error (v1.4.42) 2025-12-28 15:54:20 +08:00
Hawk Lee b97617b273 fix: remove shadowed sys import causing UnboundLocalError (v1.4.41) 2025-12-28 15:47:00 +08:00
Hawk Lee 3fcf7e4ff5 fix: mock deepspeed for resemble-enhance and silence cosyvoice logs (v1.4.40) 2025-12-28 15:35:54 +08:00