fix: only train for one batch in PyTorch Trainer test mode #8260
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
✅ Deploy Preview for determined-ui ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
Comment on lines
753
to
759
| if self.local_training: | ||
| searcher_length = self.max_length | ||
| train_length = Batch(1) if self.test_mode else self.max_length | ||
| else: | ||
| searcher_length = TrainUnit._from_searcher_unit( | ||
| train_length = TrainUnit._from_searcher_unit( | ||
| op.length, self.searcher_unit, self.global_batch_size | ||
| ) | ||
| assert searcher_length | ||
| assert train_length |
There was a problem hiding this comment.
just confirming, we don't support --test without --local through any codepath except the legacy codepath, right?
This seems correct under that condition, but might make more sense to check for test mode first, then local, since if we did support nonlocal test mode we would still only train one batch.
There was a problem hiding this comment.
good point. updated.
Comment on lines
753
to
760
| if self.local_training: | ||
| searcher_length = self.max_length | ||
| train_length = self.max_length | ||
| else: | ||
| searcher_length = TrainUnit._from_searcher_unit( | ||
| train_length = TrainUnit._from_searcher_unit( | ||
| op.length, self.searcher_unit, self.global_batch_size | ||
| ) | ||
| assert searcher_length | ||
| if self.test_mode: | ||
| train_length = Batch(1) |
There was a problem hiding this comment.
nit: reorder your if statements so you aren't mutating variables; that would feel a lot more natural:
if self.test_mode:
train_length = Batch(1)
elif self.local_training:
train_length = self.max_length
else:
train_length = TrainUnit._from_searcher_unit(...)There was a problem hiding this comment.
duh. thanks.
rb-determined-ai
approved these changes
Oct 31, 2023
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.
Description
previous change to enable training lengths longer than searcher length introduced bug where test_mode flag will train for max_length instead of just one batch.
Test Plan
run a script using pytorch trainer API (example for mnist):
verify in stdout that the code trains for 1 batch only.
Commentary (optional)
Checklist
docs/release-notes/.See Release Note for details.
Ticket