Luca207/deepseek_ocr_finetuned2
LoRA-style vs deepseek-ai/DeepSeek-OCR. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 12 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 2148/2303, Attention 48/216