vadimbelsky/qwen3-TTS-KSA
Partial fine-tune vs Qwen/Qwen3-TTS-12Hz-1.7B-Base. Partial change: 83% of the weights changed (attention, embeddings, lm_head, mlp, norms, other), in 28 of 28 layers.
Tensors changed: lm_head 14/15, Embeddings 16/17, Norms 74/134, MLP 99/103, Attention 132/132, other 3/3