KaLM-Embedding/KaLM-Reranker-V1-Large
LoRA-style vs google/t5gemma-2-4b-4b. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, q_proj, v_proj changed, in 34 of 34 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, MLP 0/312, Norms 0/521, Attention 204/488, other 0/1