minte1431/Gemma-4-12B-OBLITERATED

Partial fine-tune vs google/gemma-4-12B-it. Partial change: 48% of the weights changed (attention, embeddings, mlp), in 21 of 48 layers.

Tensors changed: Embeddings 1/1, Norms 0/291, other 0/54, MLP 63/144, Attention 80/184

22 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse