aemmeath/MiniCPM5-Aether
LoRA-style vs openbmb/MiniCPM5-2B. 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 42 of 42 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/85, MLP 126/126, Attention 168/168
1 derivatives checked
- aemmeath/MiniCPM5-Astra: LoRA-style