v5.4.0 breaks `PretrainedConfig` type checking

[fynnsu]

System Info

transformers 5.4.0
mypy 1.19.1
python 3.10

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

# transformers_typing.py
from transformers import LlamaConfig
llama_config = LlamaConfig(vocab_size=32000)

Run mypy type checker

mypy transformers_typing.py

Output:

transformers_typing.py:3: error: Unexpected keyword argument "vocab_size" for "LlamaConfig"  [call-arg]
Found 1 error in 1 file (checked 1 source file)

Note: It's not just vocab_size that's missing. As far as I can tell, this fails with all config attributes.

Expected behavior

Type checking passes. Note: this is the case for transformers<5.4.0.

[Rocketknight1]

We don't guarantee mypy compatibility unfortunately! If someone can find a clean fix that restores it, we might merge that PR, but we don't want to add a lot of code complexity to do it

[shhKnight30]

@Rocketknight1 please look at my PR once

[zucchini-nlp]

@fynnsu hey, we'll merge the suggestion from @shhKnight30

We don't guarantee that mypy would work, indeed. Instead, we are slowly moving towards making transformers ty-compatible, and I can confirm that ty complains about the same things