LunarOilRig/PaddleOCR-VL-manga-GGUF

Quantization (fewer layers) vs jzhang533/PaddleOCR-VL-For-Manga. LunarOilRig/PaddleOCR-VL-manga-GGUF matches part of jzhang533/PaddleOCR-VL-For-Manga (165 of its 620 tensors; the rest of jzhang533/PaddleOCR-VL-For-Manga has no counterpart). On the shared part: LunarOilRig/PaddleOCR-VL-manga-GGUF is a GGUF quantization (Q4_K_M) of the same weights: 37 tensors match exactly and 128 within quantization error.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/37, MLP 0/54, Attention 0/72

On the Hub · Downloads on Model Pulse