Skip to content

Conversation

@aobo-y
Copy link
Contributor

@aobo-y aobo-y commented Nov 25, 2025

Summary: if mask is None, the whole image is seen as one interpretable feature. Initiate a dummy mask whose pixels all belongs to feature id 0, so the following code can be simplified by assuming mask is given

Differential Revision: D87890030

@meta-cla meta-cla bot added the cla signed label Nov 25, 2025
@meta-codesync
Copy link
Contributor

meta-codesync bot commented Nov 25, 2025

@aobo-y has exported this pull request. If you are a Meta employee, you can view the originating Diff in D87890030.

aobo-y added a commit to aobo-y/captum that referenced this pull request Nov 25, 2025
…ch#1674)

Summary:

if `mask` is `None`, the whole `image` is seen as one interpretable feature. Initiate a dummy `mask` whose pixels all belongs to feature id `0`, so the following code can be simplified by assuming `mask` is given

Differential Revision: D87890030
Summary:

add `MMImageMaskInput`, the most naive interpretable input for image in LLM
- only support image features, not text/mixed
- only 1 image for now
- need a given `mask` for segmentation

Differential Revision: D87818563
Summary:

add tests for `MMImageMaskInput`

Differential Revision: D87822368
…ch#1674)

Summary:

if `mask` is `None`, the whole `image` is seen as one interpretable feature. Initiate a dummy `mask` whose pixels all belongs to feature id `0`, so the following code can be simplified by assuming `mask` is given

Differential Revision: D87890030
@meta-codesync
Copy link
Contributor

meta-codesync bot commented Dec 1, 2025

This pull request has been merged in ff0eaf3.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants