SIGN IN SIGN UP

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

0 0 183 Python

Untangle config inheritance (#41541)

* remove from base

* delete

* fetcher fix

* missing values

* update

* is decoder missing

* forgot to add

* add special tokens with default `None` in text models

* fsmt has unused subconfig, fix it!

* update

* fix

* add missig token id defaults

* fix more tests

* tie_word_embeddings

* tiny fixes

* more test fixes

* fix docstrings

* fix copies

* fix style?

* fix copied again

* fix copies

* fix examples

* delete left over print stmt

* splitnter

* this defi will fix a bunch decoder-only models

* make style

* fix copies

* not all models are supposed to have an attr for `tie_word_embeddings`!

* comment out

* fix

* more fixes

* fix copies

* docstring and non-model tests

* update

* fix repo consistency

* style

* fix

* revert

* remove unused attr

* fix repo

* fix test

* fix a few tests, more tests

* fix gemma & llava

* style

* gemma3n also

* new models as well

* skip the test

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
R
Raushan Turganbay committed
aec03e5d1921038bdcb0bd8915e696ca02c82983
Parent: 79f0386
Committed by GitHub <noreply@github.com> on 1/16/2026, 9:32:39 AM