[memory] Avoid storing trainer in ModelCardCallback and SentenceTransformerModelCardData
#3144
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
and SentenceTransformerModelCardData This prevents a proper cleanup
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Outdated suggestions cannot be applied.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.
Resolves #3136
Hello!
Pull Request overview
Details
This seems to prevent cleanup, as there's a cyclical dependency between trainer -> model -> model card -> trainer. This means that once the trainer and model get overridden (e.g. in https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/data_augmentation/train_sts_seed_optimization.py), the old model/trainer/model_card_data don't get automatically eaten by the garbage disposal.
I've moved a lot of components around, and now ModelCardCallback nor SentenceTransformerModelCardData need to store the Trainer. Although annoying, this does mean that memory should be cleared if the model/trainer gets overridden/deleted.
Before:
Approximate highest recorded VRAM during train_sts_seed_optimization:
After
Approximate highest recorded VRAM during train_sts_seed_optimization:
Note that the VRAM usage does still grow, albeit a lot more slowly, so this might not have resolved all issues. Having said that, because most people only make 1 trainer, it's not that big of an issue I suspect.