Python code in model docs

[zucchini-nlp]

What does this PR do?

• Deletes all torch_dtype and dtype, since we load models in default dtype from config in v5
• Uses plain python code (find and delete all >>> and ... )

Batch updating because I see many comments in new model PRs asking users to fix it, though the best way is for us to fix old docs

@vasqu , do you have any other recurring/annoying patterns? Would be great to reduce number of nitty-picky PR comments, and save us time for more important tasks

[vasqu]

One quick one: missing device map auto on models and or missing movement from prepared input .to(model.device)

[zucchini-nlp]

This will need an agent help, not super regex friedly. Oke, noted

[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.

[zucchini-nlp]

@stevhliu this PR is generated by find-replace and code agent. As per title, eliminates most common patterns in new model reviews, so users don't copy a bad example

Whenever you have time, can you reveiew?

[vasqu]

@zucchini-nlp Another one maybe for a different PR: Using httpx vs requests. Sorry dropping stuff as I review other models atm :D

[zucchini-nlp]

No prob, gotta keep docs clean and anyway it is the LLM doing the work, I just re-check. Defi another PR opportunity :)

[stevhliu]

nice, thanks for the clean up! 🧹 some issues i've noticed so far:

• using the wrong bnb arg (bnb_4bit_compute_bnb_4bit_quant_type vs bnb_4bit_quant_dtype)
• double commas after the model
• indenting off for some device_map args
• some examples use device=0 while others device=0 is removed

[zucchini-nlp]

Love vibe-coding, always have to check and clean manually afterwards 😄 I'll go over and fix those

[zucchini-nlp]

I think it is better now, applied ruff to fix indentation, reverted all device=0 which were deleted incorrectly and fixed the merged Bnb names

[stevhliu]

just a few more things, but otherwise lgtm! i'll add these conventions to my model doc skill so all future model docs follow these :)

[zucchini-nlp]

Yep, thanks, imo with the skill and a bit more requirements on using the provided doc-template, we can cut down time spent reviewing new model docs