vidore/colpali-v1.3-hf
LoRA-style vs vidore/colpaligemma-3b-pt-448-base. 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 18 of 27 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/147, MLP 54/162, Attention 72/288, other 0/2
2 other repos hold byte-for-byte identical weight files.