transformers

docs: fix OOM/hang for Qwen dynamic resolution models by setting min_pixels and max_pixels

#46007

Open