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.

Sign up for GitHub

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

Jump to bottom

Why uniform noise is scaled by 512/seq_len? #172

Open

w32zhong opened this issue Dec 31, 2024 · 0 comments

w32zhong commented Dec 31, 2024

In training script, the uniform random noise (with std=0.1 as written on paper) is additionally scaled by 512/seq_len:

noise = (torch.rand_like(tensor) - 0.5) * self.std * 512 / tensor.shape[1]

I am wondering why this is practiced, and is there any reason behind this scaling?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment