AutoModel(s) do not respect the `revision` flag while loading custom models

[ankrgyl]

System Info

• transformers version: 4.21.1
• Platform: macOS-12.4-arm64-arm-64bit
• Python version: 3.10.5
• Huggingface_hub version: 0.8.1
• PyTorch version (GPU?): 1.12.1 (False)
• Tensorflow version (GPU?): not installed (NA)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Using GPU in script?: no
• Using distributed or parallel set-up in script?:no

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

from transformers import AutoModelForImageClassification
m = AutoModelForImageClassification.from_pretrained(
        "sgugger/custom-resnet50d",
        trust_remote_code=True,
        revision="ed94a7c6247d8aedce4647f00f20de6875b5b292"
)
# It will print:
# Explicitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.

I stepped through the code and observed that AutoConfig.from_pretrained here swallows the revision from kwargs, meaning that later on line 433 it's no longer there. I believe the same issue applies to use_auth_token.

Expected behavior

I think the revision should propagate to both the configuration and model.

[LysandreJik]

cc @sgugger

[sgugger]

Thanks for flagging! The PR linked above should solve this.

[ankrgyl]

Appreciate the quick turnaround :)