-
Notifications
You must be signed in to change notification settings - Fork 322
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
RuntimeError: shape '[540672, 1]' is invalid for input of size 655360 #141
Comments
Is it caused by |
Using 8x A100s Tried those batch sizes without success:
|
Here is a minimal dataset example of 16 samples using batch_size=16, with which we are running into this issue: ds.zip |
Hmm anyone else been able get past this stage in training yet? Should be straight forward to replicate with that minimal dataset, to us right now it doesn't look like training from scratch works at all. Thanks. |
I'm away for conference now and can't help until Dec 18. I have added a label and see if anyone else could help debug. |
I've successfully trained a low-quality model from scratch with ~200 WAV files of 1 to 2,5 seconds in duration, so I can confirm that the system indeed works and this would be something more local - have you guys tried it with a small number of batches, such as 4 or 8 yet? |
Tried with batch_size=4 on 2x 4090s, running into the same issue. I recorded the training approach, in case we are doing something wrong?
|
* Update server_fastapi.py. Add new api endpoints. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Trying to do a training from scratch, experimenting with a small dataset to understand the training flow.
ZeroDivisionError
exception
full output
Division through 0 happens here:
StyleTTS2/train_second.py
Line 676 in 1ece0a3
Because in the torch loop there is an exception, so
StyleTTS2/train_second.py
Line 671 in 1ece0a3
StyleTTS2/train_second.py
Line 569 in 1ece0a3
The exception is invisible because of the try/except block:
StyleTTS2/train_second.py
Lines 672 to 673 in 1ece0a3
I added:
Which shows up the underlaying exception:
exception
full output
Note, just experimenting with minimal setup to get familar with training, therefore low # of epochs, max_len, etc
Configs/config.yml
The text was updated successfully, but these errors were encountered: