gepardzik/G4-Exp1-26B-A4B

Full fine-tune vs google/gemma-4-26B-A4B. Fine-tune: 98% of the weights changed, across 30 of 30 layers.

Tensors changed: Embeddings 2/4, MLP 150/231, Norms 211/433, other 59/122, Attention 115/223

2 derivatives checked