erkamk/qwen3-tts-tiny-tr
Full fine-tune vs Qwen/Qwen3-TTS-12Hz-0.6B-Base. Fine-tune: 92% of the weights changed, across 8 of 8 layers.
Tensors changed: other 0/74, lm_head 15/15, Norms 25/25, MLP 6/6, Attention 8/8
Full fine-tune vs Qwen/Qwen3-TTS-12Hz-0.6B-Base. Fine-tune: 92% of the weights changed, across 8 of 8 layers.
Tensors changed: other 0/74, lm_head 15/15, Norms 25/25, MLP 6/6, Attention 8/8