Destroy ZeRO #4383
Open
Destroy ZeRO #4383
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
jomayeri
requested review from
jeffra,
tjruwase,
samyam and
mrwyattii
as code owners
tjruwase
reviewed
Sep 22, 2023
tjruwase
reviewed
Sep 22, 2023
| @@ -519,6 +519,9 @@ def __init__(self, | |||
| self._enable_universal_checkpoint() | |||
| self._param_slice_mappings = self._create_param_mapping() | |||
|
|
|||
| def destroy(self): | |||
| self._uncreate_param_mapping() | |||
There was a problem hiding this comment.
Don't we need to delete the buffers related to contiguous_gradients and gradient_partitions?
There was a problem hiding this comment.
Deleting those buffers didn't make a difference when running experiments but I can add a full delete of all attributes for completeness just as I did for ZeRO 3.
There was a problem hiding this comment.
Sounds good. Is it possible to make the attribute deletion function a utility that lives in deepspeed.utils?
mrwyattii
reviewed
Sep 22, 2023
tjruwase
reviewed
Sep 24, 2023
tjruwase
reviewed
Sep 24, 2023
| @@ -519,6 +520,10 @@ def __init__(self, | |||
| self._enable_universal_checkpoint() | |||
| self._param_slice_mappings = self._create_param_mapping() | |||
|
|
|||
| def destroy(self): | |||
There was a problem hiding this comment.
This is needed in the other optimizer implementations:
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Outdated suggestions cannot be applied.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.
Create destroy function for zero stage 1 and 2, update the current function for zero 3. Add a destructor to the engine so that it releases memory at exit.