Free outputs callback #2598
Merged
Free outputs callback #2598
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
b-chu
approved these changes
Oct 3, 2023
|
Nit: isn't free a bit of an overloaded term in ML? Or in general? What about "releaseOutputs"? That's a bit more of a C++ pointer terminology here but its a bit more exact. |
Co-authored-by: Charles Tang <j316chuck@users.noreply.github.com>
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Outdated suggestions cannot be applied.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.
What does this PR do?
Adds callback to support freeing outputs for memory savings. When not using
train_metrics,self.state.outputsare not needed. However, they may take up a non-trivial amount of memory (seq_length*vocab_size*microbatch_size*bytes_per_param). For certain long sequence models, this memory starts to matter (~1-2GB), so having an option to free the memory is useful.Existing tests (eg
TestCallbackTrains) should be sufficient for this PR.What issue(s) does this change relate to?
GRT-2464