PiotrSty/trocr-pl-mixed-v3
LoRA-style vs PiotrSty/trocr-pl-base. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, out_proj, q_proj, v_proj changed, in 12 of 12 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/8, Attention 96/252, Norms 0/122, MLP 0/48, other 0/50