[Pallas] Introduce gmm_backward #7151
Merged
Conversation
This file contains hidden or 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
JackCaoG reviewed May 30, 2024
| | ||
| @unittest.skipIf(xr.device_type() != 'TPU', "This test only works on TPU.") | ||
| def test_gmm_backward(self): | ||
| self._init_test_cases() |
There was a problem hiding this comment.
nit, you might need a met.clear_all() here.
There was a problem hiding this comment.
Let me fix it in the next PR. Don't want to waste CI cycles.
JackCaoG approved these changes May 30, 2024
| Skip GPU tests to move fast. |
Just we need to override the MoE back prop to accommondate gmm backward and manual sharding. You will know what that means once the code is ready. Thanks for approving this change. |
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 Outdated suggestions cannot be applied.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.
Summary:
This pull request introduces a helper for gmm_backward. I'm still debuting if we need to make gmm as a autograd.function given we will do manual back-propagation in Mixtral.
Test Plan:
python test/test_gmm.py