curator-research/Llama-3.3-Nemotron-Cascade-2-30B-A3B-Curator-0.1.0-Enrichment-Thinking-ON

Full fine-tune vs nvidia/Nemotron-Cascade-2-30B-A3B. Fine-tune: 94% of the weights changed, across 52 of 52 layers.

Tensors changed: other 23/207, Embeddings 0/1, Norms 0/76, MLP 2688/2691, Attention 24/24, lm_head 0/1

On the Hub · Downloads on Model Pulse