Hikari07jp/DSpark-Qwen3.6-27B-AEON-draft
Full fine-tune vs z-lab/Qwen3.6-27B-DFlash. Fine-tune: 100% of the weights changed, across 5 of 5 layers.
Tensors changed: other 1/1, Norms 20/22, MLP 15/15, Attention 20/20
Full fine-tune vs z-lab/Qwen3.6-27B-DFlash. Fine-tune: 100% of the weights changed, across 5 of 5 layers.
Tensors changed: other 1/1, Norms 20/22, MLP 15/15, Attention 20/20