mhenrichsen/danskgpt-tiny-chat
Full fine-tune vs mhenrichsen/danskgpt-tiny. Fine-tune: 100% of the weights changed, across 22 of 22 layers.
Tensors changed: Norms 45/45, MLP 66/66, Attention 88/88
Full fine-tune vs mhenrichsen/danskgpt-tiny. Fine-tune: 100% of the weights changed, across 22 of 22 layers.
Tensors changed: Norms 45/45, MLP 66/66, Attention 88/88