Use general way to access and update submodules #371
Merged
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
[ghstack-poisoned]
Merged
kwen2501
added a commit
that referenced
this pull request
May 29, 2024
ghstack-source-id: 9372cca2b62078993fa7cd8f28d3b5bb710b406b Pull Request resolved: #371
facebook-github-bot
added
the
CLA Signed
[ghstack-poisoned]
kwen2501
added a commit
that referenced
this pull request
May 29, 2024
ghstack-source-id: 6eff83d7fe5af576dc6da0dcdae5bc51b4ac8ec4 Pull Request resolved: #371
wconstab
reviewed
May 29, 2024
| @@ -369,7 +369,7 @@ def parallelize_llama(model, world_mesh, parallel_dims, job_config: JobConfig): | |||
| # apply AC + torch.compile | |||
| ac_config = job_config.activation_checkpoint | |||
| enable_compile = job_config.training.compile | |||
| for layer_id, transformer_block in model.layers.items(): | |||
| for layer_id, transformer_block in model.layers.named_children(): | |||
There was a problem hiding this comment.
is there any ordering gaurantee of named_children though?
i think this will work only because the names are assumed to be '0', '1', ...
There was a problem hiding this comment.
actually, does order even matter here? i was thinking more about the model's forward code, but that code doesn't get affected by this.
This PR fixes the issue mentioned [here](pytorch/pytorch#126653 (comment)): "Module object has no attributed items." The reason is, a split `ModuleDict` is no longer a `ModuleDict`. (Future support is not guaranteed.) It would be more generally applicable if we use `named_children()` and `register_module()` to access and update submodules. [ghstack-poisoned]
wconstab
approved these changes
May 29, 2024
tianyu-l
pushed a commit
that referenced
this pull request
Aug 16, 2024
ghstack-source-id: ba1d77e5825a26632fe9b7509a88b44509cac45f Pull Request resolved: #371
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):
This PR fixes the issue mentioned here:
"Module object has no attributed items."
The reason is, a split
ModuleDictis no longer aModuleDict.It would be more generally applicable if we use
named_children()andregister_module()to access and update submodules.