schift-io/schift-ocr-1-beta
LoRA-style vs baidu/Unlimited-OCR. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, up_proj changed, in 11 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/7, other 0/12, Norms 0/171, MLP 654/2292, Attention 0/216