Skip to content

Fix bugs in cutout training #233

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.

Sign up for GitHub

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

Jump to bottom
Merged
merged 2 commits into from
May 21, 2021
Merged

Fix bugs in cutout training #233

merged 2 commits into from
May 21, 2021

Conversation

ravinkohli
Copy link
Contributor

This PR fixes the following bugs:

  1. we were not sampling the indices with replacement.
  2. we were taking the minimum between batch size and num features, which is not necessary
  3. Numerical features were getting -1 to them, whereas 0 makes more sense

@ArlindKadra ArlindKadra self-requested a review May 21, 2021 14:17
ArlindKadra
ArlindKadra reviewed May 21, 2021
View reviewed changes
autoPyTorch/pipeline/components/training/trainer/RowCutMixTrainer.py Outdated
@@ -39,7 +39,8 @@ def data_preparation(self, X: np.ndarray, y: np.ndarray,
# It is unlikely that the batch size is lower than the number of features, but
# be safe
size = min(X.shape[0], X.shape[1])
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This should also be changed to size=X.shape[1] right ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

True, I missed that

ArlindKadra
ArlindKadra reviewed May 21, 2021
View reviewed changes
autoPyTorch/pipeline/components/training/trainer/RowCutOutTrainer.py

# We use an ordinal encoder on the tabular data
if not isinstance(self.numerical_columns, typing.Iterable):
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if the numerical columns are None, we should still continue with only categorical imputing in this case or not.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also if there are only numerical columns, there should not be a conversion for categorical ones.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually when there are no numerical columns, it is not none but it is an empty list. And indexing with an empty list does not affect the tensor so this should work

Copy link

@ArlindKadra ArlindKadra May 21, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Auto-PyTorch/autoPyTorch/pipeline/components/training/trainer/base_trainer_choice.py

Lines 340 to 341 in 8b71ee2

numerical_columns=X['dataset_properties']['numerical_columns'] if 'numerical_columns' in X[
'dataset_properties'] else None

Is numerical_columns always in dataset_properties ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

when its tabular data then yeah

@ArlindKadra ArlindKadra merged commit 463c166 into refactor_development_regularization_cocktails May 21, 2021
github-actions bot pushed a commit that referenced this pull request May 21, 2021
Ravin Kohli: Fix bugs in cutout training (#233)
061fb0a
@ravinkohli ravinkohli mentioned this pull request Jun 17, 2021
Closed
@ravinkohli ravinkohli deleted the fix_cutTrainer branch October 22, 2021 09:36
ravinkohli added a commit that referenced this pull request Dec 8, 2021
@ravinkohli
Fix bugs in cutout training (#233)
a41a2a3 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Dec 8, 2021
@ravinkohli
Fix bugs in cutout training (#233)
d0f2875 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Dec 21, 2021
@ravinkohli
Fix bugs in cutout training (#233)
8b8ba42 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Jan 24, 2022
@ravinkohli
Fix bugs in cutout training (#233)
f026103 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Jan 28, 2022
@ravinkohli
Fix bugs in cutout training (#233)
d140442 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Feb 28, 2022
@ravinkohli
Fix bugs in cutout training (#233)
a831767 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Feb 28, 2022
@ravinkohli
Fix bugs in cutout training (#233)
e86fbcf 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Mar 9, 2022
@ravinkohli
Fix bugs in cutout training (#233)
17d18d8 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit to ravinkohli/Auto-PyTorch that referenced this pull request Apr 12, 2022
@ravinkohli
Fix bugs in cutout training (automl#233)
64b1397 
* Fix bugs in cutout training

* Address comments from arlind
ravinkohli added a commit that referenced this pull request Jul 26, 2022
@ravinkohli
Fix bugs in cutout training (#233)
c4b7729 
* Fix bugs in cutout training

* Address comments from arlind
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ArlindKadra ArlindKadra ArlindKadra left review comments

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@ravinkohli @ArlindKadra