Update FeatureAblation to handle precision loss when baseline is more granular than input when cross tensor attribution is enabled #1644
Conversation
This file contains hidden or 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
Contributor
| This pull request was exported from Phabricator. Differential Revision: D81980219 |
… granular than input when cross tensor attribution is enabled (meta-pytorch#1644) Summary: Noticed when flipping the flag, this test case failed: https://www.internalfb.com/code/fbsource/[faf71541b1ec0fae639f82d487b81fb18ea3e523]/fbcode/pytorch/captum/tests/attr/test_dataloader_attr.py?lines=138%2C134 The ablated tensor was `tensor([0])` instead of `tensor([[0.1])` since the baseline was a float-type and the input tensors were int tensors. https://www.internalfb.com/code/fbsource/[f2fcc926a6f3669602bac4d28c2d92e4197c96b9]/fbcode/pytorch/captum/captum/attr/_core/feature_ablation.py?lines=707-709 `ablated_input` is just a copy of the `input_tensor`, so during assignment, the ablated feature tensor incorrectly gets cast to an int tensor for this case. Differential Revision: D81980219
sarahtranfb force-pushed the export-D81980219 branch from 6b7847d to f2ab145 Compare Contributor
| This pull request was exported from Phabricator. Differential Revision: D81980219 |
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
2 participants
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 Outdated suggestions cannot be applied.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.
Summary:
Noticed when flipping the flag, this test case failed:
https://www.internalfb.com/code/fbsource/[faf71541b1ec0fae639f82d487b81fb18ea3e523]/fbcode/pytorch/captum/tests/attr/test_dataloader_attr.py?lines=138%2C134
The ablated tensor was
tensor([0])instead oftensor([[0.1])since the baseline was a float-type and the input tensors were int tensors.https://www.internalfb.com/code/fbsource/[f2fcc926a6f3669602bac4d28c2d92e4197c96b9]/fbcode/pytorch/captum/captum/attr/_core/feature_ablation.py?lines=707-709
ablated_inputis just a copy of theinput_tensor, so during assignment, the ablated feature tensor incorrectly gets cast to an int tensor for this case.Differential Revision: D81980219