yoavipo/nativ-he-decision
Full fine-tune vs dicta-il/neodictabert. Fine-tune: 98% of the weights changed, across 0 of 0 layers.
Tensors changed: other 29/29, Norms 57/57, MLP 56/56, Attention 23/28
Full fine-tune vs dicta-il/neodictabert. Fine-tune: 98% of the weights changed, across 0 of 0 layers.
Tensors changed: other 29/29, Norms 57/57, MLP 56/56, Attention 23/28