Fix MXFP4 quantizer to support variable num_local_experts and hidden_size

[marksverdhei]

What does this PR do?

This PR replaces hardcoded values num_local_experts and hidden_size in MXFP4Config for GPT-OSS type models.

I discovered this when experimenting with non-standard configs of GPT-OSS architecture but i'm pretty sure it'll break for openai/gpt-oss-120b as well since it's number of experts is different from the hardcoded value.

The quantizer hardcoded 32 experts and 2880 hidden_size in the reshape operations. This caused failures when quantizing models with different numbers of experts.

Changes:

• Read num_local_experts and hidden_size from model.config
• Use dynamic values in reshape operations instead of hardcoded constants
• Defaults to 32 and 2880 for backward compatibility

This enables quantizing averaged/merged MoE models with fewer experts.
Passed all tests that I was able to run locally on 24gb of vram.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - no
• Did you read the contributor guideline,
  Pull Request section? - yes
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case. - I looked and didn't find an issue
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings. - likely not necessary
• Did you write any new necessary tests? - no, unsure if needed

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[Rocketknight1]

cc @MekkCyber for quantization

[MekkCyber]

run-slow: mxfp4

[github-actions]

This comment contains run-slow, running the specified jobs:

models: []
quantizations: ['quantization/mxfp4'] ...

[MekkCyber]

Sounds good Thanks!

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: mxfp4

[MekkCyber]

run-slow: mxfp4

[github-actions]

This comment contains run-slow, running the specified jobs:

models: []
quantizations: ['quantization/mxfp4'] ...

[marksverdhei]

run-slow: mxfp4

Regarding the tests: Executing the custom container implementation failed. Please contact your self hosted runner administrator.
To me it looks like it failed because of issues with the CI infra. I wasn't able to see any logs in the gh actions logs.
Otherwise, I'm curious where I'm supposed to find the result of the actual pytest run

[MekkCyber]

run-slow: mxfp4

[MekkCyber]

hey @marksverdhei, yes it seems it was a docker issue, I relaunched the tests let's see if it works

[github-actions]

This comment contains run-slow, running the specified jobs:

models: []
quantizations: ['quantization/mxfp4'] ...

[MekkCyber]

seems to be working 🥳

[HuggingFaceDocBuilderDev]

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.