-
Notifications
You must be signed in to change notification settings - Fork 151
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
Multi speaker training error #69
Comments
Since you specify a gin_channel to be 12, and 204-192 is exactly 12, so I suspect the input to the proj_w layer in the encoder is not concatenated by speaker embedding somehow. You can debug into the code to see the tensor shapes. |
It came to me that you probably need to check if the speaker information is correctly fed into the model. In the code provided by the authors, speaker identity (the variable |
Thank you for the reply! ===init,py===
===train.py===
|
I don't see any problems from the code you attached, but I discover that the code in your error log (the error messages that you provided at the first of this post) differs from the |
Yes I am still getting the error. Is there any code I can refer to? Or maybe it's a CUDA or GPU issue? |
I do not think this is related to device or hardware, as CUDA/GPU won't change the data attributes (values, shapes, etc.) themselves. By the error messages, I guess this is closely related to the Line 105 in 13e9976
g or not. I think the problem is caused around there as your error message hinted. Maybe a more careful look into the code (together with tensor shapes) is necessary.
|
Thank you so much for your kind reply.
And when you see tensor shapes while debugging, you mean visualization? |
Usually this kind of errors are caused from
As for the tensor shapes, I mean the shape (dimensions) of a Pytorch tensor object. In other words, the |
I understand now about tensor shapes. The number of speakers is 12, and starting with number 1.
|
Yes, then you should begin the index with 0, because an Embedding layer of size |
Thank you so much. Learning goes on! |
Hi, I am trying to train a multi-speaker, but when I run "train.py" I get the following error:
Single-speaker training works fine, but when I add "gin_channels" and start multi-speaker training, I get an error.
computer specifications are
GPU: RTX 3090 x2
CUDA: 11.1
Pytorch: 1.7.1+cu110
How can I solve this?
The text was updated successfully, but these errors were encountered: