Bug fix: contribution renormalization correction for FFNs

[amodaresi]

While the apply_threshold_and_renormalize method works fine in renormalizing attention contribution scores, it doesn't renormalize correctly for FFNs. The reason is that when using this function for FFNs, the resid_dims and block_dims are the same. Therefore, in this line in the original code:

llm-transparency-tool/llm_transparency_tool/routes/contributions.py

Line 197 in d8e249e

denom = c_residual + c_blocks.sum(dim=tuple(range(resid_dims, block_dims)))

the sum operation receives an empty tuple for the dimensions, leading to the entire c_blocks tensor being summed and returning a single scalar value for the whole input length. This causes the returned tensors of this function not to sum up to one for each representation.

In this fix, I’ve added a condition to check if resid_dims and block_dims are the same, in which case c_blocks is added directly to c_residual to calculate denom.

[facebook-github-bot]

Hi @amodaresi!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!