transformers
Fix `base_model_prefix` in the PP-OCR detectors, CLIPSeg, TimesFM, CLIP and VideoPrism heads so base checkpoints load
#48871
Merged

Loading