bingyang-lei/Qwen3-8B-Ins-Draft-OPD
Full fine-tune vs z-lab/Qwen3-8B-DFlash-b16. Fine-tune: 100% of the weights changed, across 5 of 5 layers.
Tensors changed: other 1/1, Norms 22/22, MLP 15/15, Attention 20/20
Full fine-tune vs z-lab/Qwen3-8B-DFlash-b16. Fine-tune: 100% of the weights changed, across 5 of 5 layers.
Tensors changed: other 1/1, Norms 22/22, MLP 15/15, Attention 20/20