AIR-hl/Llama-3.2-3B-DPO

Full fine-tune vs tanliboy/llama-3.2-3b-sft. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 1/1, Norms 14/57, MLP 84/84, Attention 112/112

1 derivatives checked

• mradermacher/Llama-3.2-3B-DPO-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse