amd/DeepSeek-R1-0528-ptpc
LoRA-style vs deepseek-ai/DeepSeek-R1-0528. 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 changed, in 61 of 61 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Attention 304/305, Norms 0/245, MLP 990/1108, lm_head 0/1