-
Notifications
You must be signed in to change notification settings - Fork 24
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
[BUG] Extension hook executes on custom module #6
Comments
Hey Max, thanks for taking the time and reporting such issues here. This helps us a lot to improve the library. I think you've caught the same bug as #5. TL;DR: The problem is caused by using
Moving up the #5 on my priority list :) Would be great if you confirmed my guess. Cheers |
I see. I would suggest to remove alpha from the default configurations or to document this clearly, as this was surprising behavior to me; most people would probably not use plain SGD to train their networks. Anyways. I removed alpha:
This did not do the trick:
|
Thanks for the report, Based on the traceback, I think the problem is introduced by your custom Can you install from that branch and check if it did the trick? |
Yes, now it works, very nice! |
* [BUG] Reproduce bug described in #6 * [FIX] Skip extension hook if module has children * [FIX] flake8
Merged to |
Cockpit crashes with non descriptive error.
Description
I am trying to run cockpit on a simple network trained with MSE loss (although implemented through custom modules, not nn.Sequential). As far as I can see, no unsupported operations are involved (Linear, Sequential, Tanh, Identity).
Cockpit crashes on computing the BatchGrad extension. Using backpack to compute the batch gradient works without crashes.
Steps to Reproduce
Source or Possible Fix
Stacktrace
The text was updated successfully, but these errors were encountered: