BAAI/BGE-VL-v1.5-zs
LoRA-style vs llava-hf/llava-v1.6-mistral-7b-hf. 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 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: other 0/5, Embeddings 0/4, Norms 0/165, MLP 96/192, Attention 128/320, lm_head 0/1