distil-labs/distil-home-assistant-functiongemma
LoRA-style vs google/functiongemma-270m-it. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of q_proj, v_proj changed, in 18 of 18 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/109, MLP 0/54, Attention 36/72