Feature/sg 541 auto batch selection #628
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
…_batch_selection # Conflicts: # src/super_gradients/training/params.py # src/super_gradients/training/sg_trainer/sg_trainer.py
BloodAxe
reviewed
Jan 16, 2023
src/super_gradients/training/pre_launch_callbacks/pre_launch_callbacks.py
Show resolved
Hide resolved
ofrimasad
reviewed
Jan 19, 2023
src/super_gradients/training/pre_launch_callbacks/pre_launch_callbacks.py
Show resolved
Hide resolved
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Outdated suggestions cannot be applied.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.
Auto batch size selection, works out of the box for DDP.
Implemented as a "pre_launch_callback"- new feature needed for this design- upon calling train_from_config, there's often the need to modify the recipe config (for example to scale lr according to the actually number of available gpus).
The callback takes the cfg as an input, and increasingly tries different batch sizes until reaching COOM (or predefined max_batch_size), then optionally adapts the lr.
Additionally, I added the option to not kill the DDP group (needed for the consecutive batch size trial runs).