erikku-sama/lfm2-electronics-vlm-v2
LoRA-style vs LiquidAI/LFM2.5-VL-450M. 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 14 of 16 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, other 0/34, MLP 0/96, Norms 0/94, Attention 54/120