brokencircuitranch/gemma4-hermes-tools

Partial fine-tune vs google/gemma-4-26B-A4B-it. Partial change: 9% of the weights changed (attention, mlp), in 30 of 30 layers.

Tensors changed: Embeddings 0/4, MLP 93/231, Norms 0/433, other 0/122, Attention 115/223

On the Hub · Downloads on Model Pulse