change metadata key from finetuned_from to base_model

[julien-c]

related to:

• change metadata key from finetuned_from to base_model hub-docs#1084
• https://github.com/huggingface/moon-landing/issues/7584
• https://github.com/huggingface/moon-landing/pull/7350
• https://github.com/huggingface/moon-landing/pull/7989

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[Wauplin]

Looks good! I don't expect it to break anything in the wild.

[julien-c]

@LysandreJik had a good question: is the model card template used from transformers? (which could be a breaking change)

i don't remember

[julien-c]

(context: huggingface/transformers#24528 (comment))

[Wauplin]

@LysandreJik had a good question: is the model card template used from transformers? (which could be a breaking change)

I'm not aware of such a use no. AFAIK only skops and diffusers uses the template programmatically and both of them don't have a finetune_from/base_model argument.

---

For transformers, I found this file modelcard.py that implements a "model card builder" from scratch. Good reminder that we should most probably align on this.

[LysandreJik]

cc @muellerzr on aligning on this between the two libraries eventually

[julien-c]

For diffusers, it's weird that they don't use a base_model, they're supposed to output it in their model card no? cc @patrickvonplaten maybe

[Wauplin]

If it can help, I was mentioning this part of the code: https://github.com/huggingface/diffusers/blob/64603389da01082055a901f2883c4810d1144edb/src/diffusers/utils/hub_utils.py#L125.

[julien-c]

maybe raise this question in a diffusers slack channel

[Wauplin]

(link to -internal- thread)

[davanstrien]

LGTM