mlx-community/gemma-3n-E4B-it-bf16

Full fine-tune vs google/gemma-3n-E4B. Fine-tune: 100% of the weights changed, across 35 of 35 layers.

Tensors changed: MLP 108/108, Embeddings 5/5, other 281/281, Norms 247/317, Attention 140/140

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

On the Hub · Downloads on Model Pulse