Ecosyste.ms: Timeline
Browse the timeline of events for every public repo on GitHub. Data updated hourly from GH Archive.
PierrunoYT opened an issue on SWivid/F5-TTS
Unknown issue
Liangzi-qyh created a comment on an issue on SWivid/F5-TTS
Hello, I encountered the same problem as you, because the cloud server cannot directly access HuggingFace, so when executing the python gradio_app.py, it will report the same error as you, what do ...
ajkessel closed an issue on SWivid/F5-TTS
RuntimeError: Input type (c10::Half) and bias type (float) should be the same
I just started getting this error with the latest build: ``` File "/home/adam/ttspod/.venv/lib/python3.11/site-packages/ttspod/speech/f5.py", line 214, in infer_batch generated_wave = self...
TanvirHafiz opened an issue on SWivid/F5-TTS
Ran once, then exits every time
i was able to run the gradio once. it downloaded a bunch of stuff. but didnt do any conversion. tried it again and it failed to load the UI, gave this error Starting app... Running on local URL...
SWivid pushed 2 commits to main SWivid/F5-TTS
lpscr created a comment on an issue on SWivid/F5-TTS
ok i close this now @SWivid thank you very much for the reply
lpscr closed an issue on SWivid/F5-TTS
train checkpoint it's very big 5G eveytime how can make smaler in train
@SWivid i wonder if i can change this part checkpoint if curse something if i do ? https://github.com/SWivid/F5-TTS/blob/99190a03cb473f1d139b7f282e681bbd60c7fd0e/model/trainer.py#L119-L135 Wh...
SWivid created a comment on an issue on SWivid/F5-TTS
> if i have only ema can i finetune the model yes for sure, will just initialize the optimizer and scheduler and start from step=0, while loading pretrained model weights (ema basically)
SWivid created a comment on an issue on SWivid/F5-TTS
> when train finish never save the last_model ah yes, that a good point! will add a saving option before quit https://github.com/SWivid/F5-TTS/blob/99190a03cb473f1d139b7f282e681bbd60c7fd0e/mod...
lpscr created a comment on an issue on SWivid/F5-TTS
> This might cause some problems e.g. someone may need previous ckpt with optimizer and scheduler states if he/she finds that training went wild and wants to go back to some specific point. > > ...