transformers

41db33d8 - Speed up image preprocessing for vision-language models (#47453)

Speed up image preprocessing for vision-language models (#47453) * Speed up image preprocessing for vision-language models Avoid materializing repeated temporal copies when packing static image patches in GLM4V, GLM46V, GLMGA, MiniMaxM3-VL, and PaddleOCR-VL. Skip the final concatenation for single-image inputs across related image processors, and regenerate files derived from modular sources. The optimized paths preserve the existing patch order and produce bitwise-identical outputs. * Apply repo consistency fixes --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>