[FSDP][state_dict] Return tensors instead of FlatParameters to avoid pickling errors #94637
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
…pickling errors After #88913, user-defined parameter states will be pickled. For a FlatParameter, this means `_local_shard` will also be pickled. Since state_dict and load_state_dict only require the tensor, returning the full FlatParameter does not give us any extra benefit. This PR changes the behavior to simply return a view of the FlatParameter. Differential Revision: [D43205127](https://our.internmc.facebook.com/intern/diff/D43205127/) [ghstack-poisoned]
fegin
requested review from
H-Huang,
awgu,
kwen2501,
mrshenli,
rohan-varma,
wanchaol and
zhaojuanmao
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/94637
Note: Links to docs will display an error until the docs builds have been completed. ❌ 5 FailuresAs of commit 8b2f0e3: NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
the
release notes: distributed (fsdp)
rohan-varma
approved these changes
Feb 10, 2023
| if valid_data_size > 0: | ||
| if flat_param._shard_numel_padded > 0: | ||
| flat_param = flat_param.narrow(0, 0, valid_data_size) | ||
| flat_param = flat_param.view(valid_data_size) |
There was a problem hiding this comment.
nice, maybe add a comment saying this will make it return a tensor that can be properly serialized?
…s to avoid pickling errors" After #88913, user-defined parameter states will be pickled. For a FlatParameter, this means `_local_shard` will also be pickled. Since state_dict and load_state_dict only require the tensor, returning the full FlatParameter does not give us any extra benefit. This PR changes the behavior to simply return a view of the FlatParameter. Differential Revision: [D43205127](https://our.internmc.facebook.com/intern/diff/D43205127/) [ghstack-poisoned]
fegin
added a commit
that referenced
this pull request
Feb 12, 2023
…pickling errors Pull Request resolved: #94637 After #88913, user-defined parameter states will be pickled. For a FlatParameter, this means `_local_shard` will also be pickled. Since state_dict and load_state_dict only require the tensor, returning the full FlatParameter does not give us any extra benefit. This PR changes the behavior to simply return a view of the FlatParameter. ghstack-source-id: 179983735 Differential Revision: [D43205127](https://our.internmc.facebook.com/intern/diff/D43205127/)
|
@pytorchbot merge -f "The failing tests are not related." |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
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.
Stack from ghstack (oldest at bottom):
After #88913, user-defined parameter states will be pickled. For a FlatParameter, this means
_local_shardwill also be pickled. Since state_dict and load_state_dict only require the tensor, returning the full FlatParameter does not give us any extra benefit. This PR changes the behavior to simply return a view of the FlatParameter.Differential Revision: D43205127