ysober/qwen3.8-27b-dspark-zh
Full fine-tune vs RadixArk/Qwen3.8-27B-DSpark. Fine-tune: 100% of the weights changed, across 5 of 5 layers.
Tensors changed: other 5/5, Norms 22/22, Attention 10/10
Full fine-tune vs RadixArk/Qwen3.8-27B-DSpark. Fine-tune: 100% of the weights changed, across 5 of 5 layers.
Tensors changed: other 5/5, Norms 22/22, Attention 10/10