transformers

b19834d8 - fix(qwen2vl): make Qwen2VLImageProcessor respect config['size'], align with fast version (#43095)

fix(qwen2vl): make Qwen2VLImageProcessor respect config['size'], align with fast version (#43095) * fix(qwen2vl): make Qwen2VLImageProcessor respect config['size'], align with Fast version\n\n修复：https://github.com/huggingface/transformers/issues/42910\nQwen2VLImageProcessor 不再覆盖 config["size"]，一致性对齐 Fast 处理，确保 Qwen3VL 正常。 * fix(imageprocessor): make PaddleOCRVL, Video-LLaMA3 ImageProcessor respect config['size']\n\n和Qwen2VL处理方式保持一致，避免覆盖用户model config设定，提高一致性。参考 #42910