-
Notifications
You must be signed in to change notification settings - Fork 25.7k
[DCP][state_dict][doc] Update the distributed state_dict document #121290
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
As title [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/121290
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit ade614a with merge base 34a28f0 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
As title cc LucasLLC [ghstack-poisoned]
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for adding this, great explanation. One small nit
|
|
||
| To tackle these challenges, we offer a collection of APIs for users to easily manage state_dicts. `get_model_state_dict` returns a model state dictionary with keys consistent with those returned by the unparallelized model state dictionary. Similarly, `get_optimizer_state_dict` provides the optimizer state dictionary with keys uniform across all parallelisms applied. To achieve this consistency, `get_optimizer_state_dict` converts parameter IDs to fully qualified names identical to those found in the unparallelized model state dictionary. | ||
|
|
||
| Note that results returned by hese APIs can be used directly with the `save()` and `load()` methods without requiring any additional conversions. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should we clarify here that this is distributed.checkpoint.save/load vs torch?
…ocument" As title cc LucasLLC [ghstack-poisoned]
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Root may not exist due to FSDP lazy initialization. Pull Request resolved: #121544 Approved by: https://github.com/Skylion007 ghstack dependencies: #121273, #121276, #121290
Root may not exist due to FSDP lazy initialization. Pull Request resolved: pytorch#121544 Approved by: https://github.com/Skylion007 ghstack dependencies: pytorch#121273, pytorch#121276, pytorch#121290
Root may not exist due to FSDP lazy initialization. Pull Request resolved: #121544 Approved by: https://github.com/Skylion007 ghstack dependencies: #121273, #121276, #121290 Co-authored-by: Chien-Chin Huang <chienchin@fb.com>
Stack from ghstack (oldest at bottom):
As title
cc @LucasLLC