lemon-mint/gemma-2b-diff-model
Full fine-tune vs google/gemma-1.1-2b-it. Fine-tune: 100% of the weights changed, across 18 of 18 layers.
Tensors changed: Embeddings 1/1, Norms 37/37, MLP 54/54, Attention 72/72
Full fine-tune vs google/gemma-1.1-2b-it. Fine-tune: 100% of the weights changed, across 18 of 18 layers.
Tensors changed: Embeddings 1/1, Norms 37/37, MLP 54/54, Attention 72/72