Hcompany/Holotron-3-Nano

Partial fine-tune vs nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16. Partial change: 89% of the weights changed (attention, embeddings, lm_head, mlp, norms, other), in 52 of 52 layers.

Tensors changed: Norms 66/565, Embeddings 1/3, other 177/294, MLP 1587/1812, Attention 24/320, lm_head 1/1

On the Hub · Downloads on Model Pulse