Skip to content

[WIP] Support transformers weight conversion#2995

Closed
BenjaminBossan wants to merge 2 commits intohuggingface:mainfrom
BenjaminBossan:transformers-weight-conversion
Closed

[WIP] Support transformers weight conversion#2995
BenjaminBossan wants to merge 2 commits intohuggingface:mainfrom
BenjaminBossan:transformers-weight-conversion

Conversation

@BenjaminBossan
Copy link
Copy Markdown
Member

@HuggingFaceDocBuilderDev
Copy link
Copy Markdown

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

Comment on lines +269 to +270
# if target_modules is a string (=> regex), the required conversion is not trivial, we don't deal with this case
# for now
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

For the string case it is possible to load the model, get all lora layer names and therefore convert the regex to a list of names

@github-actions
Copy link
Copy Markdown

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

@BenjaminBossan
Copy link
Copy Markdown
Member Author

not stale

githubnemo pushed a commit to githubnemo/peft that referenced this pull request Feb 27, 2026
Continuation of PR huggingface#2995.
Background: huggingface/transformers#42491 and huggingface/transformers#43261.

This change implements conversion operations for converting some existing
PEFT checkpoints, mainly dealing with the fusing of MoE layers in transformers v5.

The code added here is currently a copy from the code that exists in transformers
which is supposed to be gated as soon PEFT v0.19 is released and use the code
in this PR.

The copying makes testing a bit difficult since there's currently no routing
depending on the PEFT version in transformers. Older transformers versions, therefore,
need patching to forcefully use the PEFT implementation of the conversion.
As soon as the routing is implemented in transformers we can conditionally
disable the patching.
@BenjaminBossan
Copy link
Copy Markdown
Member Author

Superseded by #3071.

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.

3 participants