This repository has been archived by the owner on Dec 16, 2022. It is now read-only.
Gradient accumulation with the distributed trainer #3537
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
brendan-ai2
reviewed
Dec 17, 2019
allennlp/training/trainer.py
Outdated
| @@ -343,21 +353,20 @@ def _train_epoch(self, epoch: int) -> Dict[str, float]: | |||
| train_generator_tqdm = train_generator | |||
|
|
|||
| cumulative_batch_size = 0 | |||
| for batch in train_generator_tqdm: | |||
| for batches in lazy_groups_of(train_generator_tqdm, self._num_gradient_accumulation_steps): | |||
There was a problem hiding this comment.
Nit: Maybe accumulation_groups (or batch_group or whatever) instead of batches? Might be confusing to a reader see lines like cur_batch = sum(training_util.get_batch_size(batch) for batch in batches).
allennlp/training/trainer.py
Outdated
| @@ -323,7 +331,9 @@ def _train_epoch(self, epoch: int) -> Dict[str, float]: | |||
|
|
|||
| # Get tqdm for the training batches | |||
| train_generator = self.iterator(self.train_data, num_epochs=1, shuffle=self.shuffle) | |||
There was a problem hiding this comment.
The lazy_groups_of call should be here to create a new generator and tqdm should be applied to that. As in
allennlp/allennlp/training/trainer.py
Line 306 in 3dda5ac
|
@brendan-ai2, I took some license with the naming, and fixed the |
Closed
This was referenced Dec 17, 2019
Closed
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
I re-did gradient accumulation with the distributed trainer.