Updating how we load metrics in a state_dict so we don't add extra memory overhead #1892
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Outdated suggestions cannot be applied.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?
When loading composer state dicts, we load in train and eval metrics. However, these metrics have a defined
_device
which is usually always the rank 0 device since that's what stored in thestate_dict.
As a result, we need to manually copy over each metric (so we don't override a user's current metrics) and manually set the devices.Before our memory overhead was:
| 0 N/A N/A 3770555 C 31567MiB |
| 0 N/A N/A 3770556 C 965MiB |
| 1 N/A N/A 3770556 C 31567MiB
Now, our memory overhead when loading models is normal:
| 0 N/A N/A 3770555 C 31567MiB |
| 1 N/A N/A 3770556 C 31567MiB |
What issue(s) does this change relate to?
CO-1485
Before submitting
pre-commit
on your change? (see thepre-commit
section of prerequisites)