RASMUS/Ahma-3B-dpo-slerp

Full fine-tune vs Finnish-NLP/Ahma-3B-Instruct. Fine-tune: 97% of the weights changed, across 26 of 26 layers.

Tensors changed: Embeddings 1/1, Norms 0/53, MLP 75/78, Attention 100/104, lm_head 1/1

1 derivatives checked

• mradermacher/Ahma-3B-dpo-slerp-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse