Fix: Allow img_size to be int or tuple in PatchEmbed

[sddongxh]

This PR modifies the PatchEmbed constructor to support img_size as either an int or a Tuple[int, int], making it compatible with downstream inputs that may provide HxW dimensions. This improves flexibility and avoids unnecessary conversion and lint error elsewhere.

Changes:

• Updated type hint for img_size to Union[int, Tuple[int, int]]
• Ensured internal usage converts int to (int, int)

No changes in behavior if used with default integer input.

[rwightman]

@sddongxh it already is fully compatible with tuple inputs, just the typing was never updated....

[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.

[sddongxh]

@sddongxh it already is fully compatible with tuple inputs, just the typing was never updated....

why not ... the lint error is quite annoying

[rwightman]

yeah, I'll merge it, the comment is suggesting that funcitonality was changed/fixed, but it's just type annotation and lint