fix: Batchnorm does not work properly when training on multiple devices #1167
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
burn-core/src/nn/norm/batch.rs
Outdated
Comment on lines
139
to
145
| let running_mean = running_mean.clone().mul_scalar(1.0 - self.momentum).add( | ||
| mean.clone() | ||
| .to_device(&running_mean.device()) | ||
| .detach() | ||
| .mul_scalar(self.momentum) | ||
| .reshape([channels]), | ||
| ); |
There was a problem hiding this comment.
I think we should do the opposite: change the device of the running_mean after detaching it so that the gradients are calculated on each device. When we store the newly updated running_mean, we should also check the devices. Do you think it would work?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
It seems that
running_varandrunninng_meanare shared across devices when training on multiple devices. Therefore, ensuring thatrunning_var/varandrunning_mean/meanare on the same device will allow Batchnorm to work properly during multiple devices training.