Skip to content

Remove instances where model_args and model_kwargs are provided #798#1424

Open
dylanberens wants to merge 3 commits into
TransformerLensOrg:dev-4.xfrom
dylanberens:add-glm4-moe-adapter
Open

Remove instances where model_args and model_kwargs are provided #798#1424
dylanberens wants to merge 3 commits into
TransformerLensOrg:dev-4.xfrom
dylanberens:add-glm4-moe-adapter

Conversation

@dylanberens

Copy link
Copy Markdown

Description

Removed *model_args from 'run_with_cache()' and 'run_with_hooks()' functions, and replaced all those function calls' positional arguments with keyword arguments-- as input= (and inputs= for all audio model calls)

Fixes #798

Type of change

Please delete options that are not relevant.

  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • This change requires a documentation update

Screenshots

N/A

Checklist:

  • I have commented my code, particularly in hard-to-understand areas
  • I have made corresponding changes to the documentation
  • My changes generate no new warnings
  • I have added tests that prove my fix is effective or that my feature works
  • New and existing unit tests pass locally with my changes
  • I have not rewritten tests relating to key interfaces which would affect backward compatibility

@jlarson4

Copy link
Copy Markdown
Collaborator

Hi @dylanberens! A couple notes on this one:

  1. Please update the PR title to match the contents of the PR
  2. Please rebase the PR to dev-4.x. This specific issue is slated for the next major version release, due to the significant nature of the breaking change
  3. Review all failing CI paths and resolve any errors. If you get to a point where all your errors are 429 errors from HuggingFace or stderr outputs in notebooks, let me know and I can manually rerun them.

Additionally, the intention is to deprecate the HookedTransformers system in 4.x. Until I get around to prepping for that release and auditing to determine which features are covered under TransformerBridge, I won't know for 100% certain what will be dropped and what will be kept. Some of your work may end up deprecated before a release happens, so if you want to wait to do this work until I have a chance to do a feature audit, that is completely acceptable.

@dylanberens dylanberens changed the title Add glm4 moe adapter Remove instances where model_args and model_kwargs are provided #798 Jun 27, 2026
@dylanberens dylanberens changed the base branch from main to dev-4.x June 27, 2026 20:32
@dylanberens

Copy link
Copy Markdown
Author

Hi @dylanberens! A couple notes on this one:

  1. Please update the PR title to match the contents of the PR
  2. Please rebase the PR to dev-4.x. This specific issue is slated for the next major version release, due to the significant nature of the breaking change
  3. Review all failing CI paths and resolve any errors. If you get to a point where all your errors are 429 errors from HuggingFace or stderr outputs in notebooks, let me know and I can manually rerun them.

Additionally, the intention is to deprecate the HookedTransformers system in 4.x. Until I get around to prepping for that release and auditing to determine which features are covered under TransformerBridge, I won't know for 100% certain what will be dropped and what will be kept. Some of your work may end up deprecated before a release happens, so if you want to wait to do this work until I have a chance to do a feature audit, that is completely acceptable.

hi @jlarson4! i've updated the title and rebased the PR to dev-4.x. I'll put this on pause until your feature audit makes it clearer what we need to adjust. thanks!

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.

[Proposal] Remove instances where model_args and model_kwargs are provided

3 participants