rawsh/mirrorqwen2.5-0.5b-ORPO-2

Partial fine-tune vs rawsh/mirrorqwen2.5-0.5b-ORPO-1. Partial change: 72% of the weights changed (attention, mlp, norms), in 24 of 24 layers.

Tensors changed: Embeddings 0/1, Norms 11/49, MLP 72/72, Attention 158/168

1 derivatives checked

• mradermacher/mirrorqwen2.5-0.5b-ORPO-2-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse