boyuzhuGPT/Qwen2.5-Omni-7B-Safety-ZeroAudio
Partial fine-tune vs Qwen/Qwen2.5-Omni-7B. Partial change: 68% of the weights changed (attention, embeddings, lm_head, mlp, norms), in 4 of 32 layers.
Tensors changed: other 0/549, MLP 84/480, Norms 57/347, Attention 196/1020, Embeddings 1/51, lm_head 1/1