SupraLabs/Supra2-Medium-Instruct
Full fine-tune vs SupraLabs/Supra2-Medium-Base. Fine-tune: 100% of the weights changed, across 7 of 7 layers.
Tensors changed: Norms 29/29, MLP 21/21, Attention 28/28
Full fine-tune vs SupraLabs/Supra2-Medium-Base. Fine-tune: 100% of the weights changed, across 7 of 7 layers.
Tensors changed: Norms 29/29, MLP 21/21, Attention 28/28