DireDreadlord/Chaturanga-Mini
Full fine-tune vs SupraLabs/Supra-50M-Base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 1/1, Norms 25/25, MLP 36/36, Attention 48/48
Full fine-tune vs SupraLabs/Supra-50M-Base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 1/1, Norms 25/25, MLP 36/36, Attention 48/48