0ppxnhximxr/Motif-3-FP8

Full fine-tune vs Motif-Technologies/Motif-3. Fine-tune: 98% of the weights changed, across 51 of 53 layers.

Tensors changed: Embeddings 0/2, Norms 0/324, Attention 0/855, MLP 102/951, other 0/103, lm_head 0/1

On the Hub · Downloads on Model Pulse