Sakalti/ultiima-72B-v2.0

Full fine-tune vs Sakalti/ultiima-72B. Fine-tune: 98% of the weights changed, across 80 of 80 layers.

Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 88/161, MLP 240/240, Attention 560/560

2 derivatives checked

On the Hub · Downloads on Model Pulse