systalyze/gemma4-26b-a4b-it-dspark

Partial fine-tune vs google/gemma-4-26B-A4B-it. Partial change: 11% of the weights changed (mlp, norms, other), in 5 of 5 layers.

Tensors changed: Embeddings 0/1, Norms 21/21, other 5/5, MLP 15/15

On the Hub · Downloads on Model Pulse