Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Reproducibility & new methods, settings, and settings #59

Merged
merged 37 commits into from
Dec 24, 2024
Merged

Conversation

loribonna
Copy link
Collaborator

Introduction of REPRODUCIBILITY.md, to keep track of which methods we have already a configuration available that reproduce the results of the original papers. Several methods have been checked, with small refactors and updates to support the setup used in their original works.

For example, for iCaRL we reproduced the original ResNet32 with the identity shortcut.

To reproduce the results of a method, check if it is supported in REPRODUCIBILITY.md and then run:
python main.py --model <model_name> --dataset <dataset_name> --model_config best [--buffer_size <buffer_size>].
For more information, visit the documentation.

Included new datasets and models:

  • added support for datasets with bias (seq-celeba)
  • added Learning without Shortcut (LwS)
  • created several configurations for datasets and models

Improved the documentations with new info and some tutorials.

@loribonna loribonna merged commit 1bca571 into master Dec 24, 2024
2 checks passed
@loribonna loribonna deleted the reproduce branch December 24, 2024 21:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant