cjvt/GaMS3-12B-Instruct-Lex
Full fine-tune vs cjvt/GaMS3-12B. Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 153/289, MLP 144/144, Attention 192/192
Full fine-tune vs cjvt/GaMS3-12B. Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 153/289, MLP 144/144, Attention 192/192