amd/DeepSeek-V3.2-mtp-ptpc
LoRA-style vs deepseek-ai/DeepSeek-V3.2. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, kv_a_proj_with_mqa, kv_b_proj, o_proj, q_a_proj, q_b_proj, up_proj, wk, wq_b changed, in 62 of 62 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/2, Attention 430/496, Norms 0/376, MLP 758/889, lm_head 0/1, other 0/2