SWivid
c3d415e47a
merging into one infer_batch_process function
2025-02-21 21:41:19 +08:00
SWivid
7ee55d773c
formatting
2025-02-21 17:00:51 +08:00
kunci115
d68b1f304c
[add] new line after gc.collect()
2025-02-21 14:48:58 +07:00
kunci115
7c0eafe240
[add] client use on readme
2025-02-21 14:45:09 +07:00
rino
4ceba6dc24
This patch is to solve a problem where streaming will handle all of the client input
...
[add] numpy tokenizer for stream chunk
[add] infer_batch_process_stream in utils_infer
[add] file writter after streaming
[edit] adjustment for streaming server
[edit] data handling processes and sends chunk by chunk
[delete] threading on processing the inference, just for file writting
2025-02-21 14:35:01 +07:00
SWivid
832ecf40b9
formatting, update readme
2025-02-19 08:35:13 +08:00
Phlo
3342859c04
feat: Add autolaunch option to Gradio interface
2025-02-18 12:29:21 -06:00
Phlo
f005f1565e
fix: typo in MPS PyTorch env variable
2025-02-18 03:28:44 -06:00
mrfakename
71ad071c1e
Update Gradio app
2025-02-14 12:44:52 -08:00
Hasan Can Solakoğlu
eebe337625
Increase batch size for text conversion from 32 to 100
2025-02-07 22:40:16 +03:00
Hasan Can Solakoğlu
0291ac17d2
Fix code formatting
2025-02-07 22:37:00 +03:00
Hasan Can Solakoğlu
bec4ebcae5
Enhance CSV preparation script to preserve order of processed audio files in chunk submissions
2025-02-07 22:35:30 +03:00
Hasan Can Solakoğlu
a9d6509a06
Enhance CSV preparation script with customizable worker count and improved usage examples
2025-02-07 22:32:42 +03:00
Hasan Can Solakoğlu
e7496d0170
Enhance audio processing with concurrent execution and graceful shutdown handling
2025-02-07 22:13:13 +03:00
Hasan Can Solakoğlu
34d94af2a8
Enhance audio duration extraction with ffprobe fallback and error handling
2025-02-07 20:38:42 +03:00
Can
33e865120c
Refactor imports and improve code formatting in dataset and trainer modules
2025-02-04 22:20:42 +03:00
Can
93ae7d3fc8
Enhance DynamicBatchSampler to support epoch-based shuffling
2025-02-04 20:21:59 +03:00
Hasan Can
bebbfbb916
Fix for incorrect defaults in the finetune_gradio interface ( #755 )
...
* Add missing components to setup_load_settings in finetune_gradio
2025-01-29 17:25:22 +08:00
unknown
f0996492a7
0.5.0 fix grad_accum bug from 0.4.0, #715 #728
2025-01-29 15:18:02 +08:00
unknown
0d95df4a4d
0.4.6 minor fixes for finetune-gradio -cli
2025-01-29 00:06:10 +08:00
Hasan Can Solakoğlu
f8cc2446c8
Fix for the checkpoint dropdown menu
2025-01-28 15:25:14 +03:00
unknown
607b92b391
0.4.5 fix extremely short case that lengths of text_seq > audio_seq, causing wrong cond_mask
2025-01-28 12:38:16 +08:00
unknown
ee2b77064e
0.4.4 fix hard coded stdout for finetune-gradio gui
2025-01-28 11:39:54 +08:00
Yushen CHEN
1e7d6da992
Merge pull request #746 from mwzkhalil/patch-1
...
Update finetune_gradio.py, set weights_only=True
2025-01-27 21:12:14 +08:00
Yushen CHEN
c2cf31e0c5
Merge pull request #729 from hcsolakoglu/fix-ckpt-rotation
...
Exclude pretrained models from the checkpoint rotation logic
2025-01-27 19:57:05 +08:00
Yushen CHEN
46266f1d14
Merge pull request #741 from Chiyan200/main
...
Fix Settings Loader Issues: Resolve KeyErrors, Path Handling, and Component Assignment (#731 )
2025-01-27 19:28:22 +08:00
mahwiz khalil
c54f4e7fc0
Update finetune_gradio.py
...
The safest approach here is to explicitly set weights_only=True to load only the model weights and avoid executing potentially unsafe code
2025-01-24 00:31:53 -08:00
98440
964064094a
Added intel XPU support
2025-01-22 03:36:10 +08:00
[Chiyan200]
24fe39dc3c
Fix : Settings Loader Issues: Resolve KeyErrors, Path Handling, and Component Assignment ( #731 )
2025-01-22 00:07:34 +05:30
[Chiyan200]
a74d0d0f83
Fix: Robust settings loader to handle missing keys, incorrect file paths, and dynamic assignment
...
- Ensured default settings are properly merged with file-based settings to prevent KeyErrors.
- Added logic to handle _pinyin and _char suffixes in project names, ensuring correct file paths.
- Implemented tuple-based ordered mapping for consistent and error-free component assignment.
- Added safety check to verify the existence of setting.json before loading.
- Improved maintainability by centralizing default settings and enhancing error handling.
2025-01-21 23:17:59 +05:30
Hasan Can Solakoğlu
2d27d2c1b2
Exclude pretrained models from the checkpoint rotation logic
2025-01-17 19:35:19 +03:00
SWivid
9e51878d18
0.4.2 fix trainer with grad_accum
2025-01-15 18:28:41 +08:00
SWivid
12d6970271
0.4.1 #718 add keep_last_n_checkpoints option
2025-01-15 15:06:55 +08:00
Hasan Can Solakoğlu
76b1b03c4d
Keep Last N Checkpoints ( #718 )
...
* Add checkpoint management feature
- Introduced `keep_last_n_checkpoints` parameter in configuration and training scripts to manage the number of recent checkpoints retained.
- Updated `finetune_cli.py`, `finetune_gradio.py`, and `trainer.py` to support this new parameter.
- Implemented logic to remove older checkpoints beyond the specified limit during training.
- Adjusted settings loading and saving to include the new checkpoint management option.
This enhancement improves the training process by preventing excessive storage usage from old checkpoints.
2025-01-15 12:28:54 +08:00
unknown
0b11f7eae6
0.4.0 fix gradient accumulation; change checkpointing logic to per_updates
2025-01-12 21:26:57 +08:00
SWivid
4872afef9f
fix no_ref_audio in cfm.py
2025-01-07 22:01:43 +08:00
hndrbrm
dc2d2d3b2f
Bug Fix: Parsing Argument --finetune always True ( #676 )
...
* Bug Fix: Parsing Argument --finetune always True
* Update finetune_cli.py, fix also log_samples & bnb_optimizer flags
* Update finetune_gradio.py. fix flags
* Update finetune_gradio.py. formatting
---------
Co-authored-by: Yushen CHEN <45333109+SWivid@users.noreply.github.com >
2024-12-29 21:29:50 +08:00
Benjamin Doron
91538ab206
utils_infer.py: Fix pathing issue that prevents importing BigVGAN
...
Upwards directory traversal goes after getting the base path.
Tested on Linux, using F5-TTS and BigVGAN. Audio is generated.
Signed-off-by: Benjamin Doron <benjamin.doron00@gmail.com >
2024-12-21 20:52:44 -05:00
SWivid
156032ac68
formatting
2024-12-21 23:20:53 +08:00
SWivid
6ab873fb19
Fixed #658
2024-12-21 23:19:11 +08:00
SWivid
deaca8d24c
v0.3.2 add flags and default values to socket_server.py
2024-12-18 20:32:20 +08:00
Jarod Mica
e278836f87
Add updated JA model
2024-12-17 14:11:32 -08:00
HotDro4illa
6cd1f9f243
fix
2024-12-17 22:15:36 +03:00
HotDro4illa
50b5e49bd6
update card
2024-12-17 22:13:43 +03:00
SWivid
9a18bbe823
v0.3.1 fix multi-style gradio bug; add features suggested #591
2024-12-17 07:57:33 +08:00
Yushen CHEN
81f69e57b7
Update infer_cli.py; typo
2024-12-16 16:31:23 +08:00
SWivid
17087321f5
formatting
2024-12-16 16:26:21 +08:00
SWivid
c93462fc68
feature. allow custom model config for gradio infer
2024-12-16 16:25:23 +08:00
ZhikangNiu
7c84e91a00
update dit checkpoint_activations and fix#399 #400
2024-12-16 13:24:58 +08:00
SWivid
c6e96d0c83
Fixed #631
2024-12-15 23:21:00 +08:00
