Vortex5/Radiant-Shadow-12B

Partial fine-tune vs Retreatcost/KansenSakura-Radiance-RP-12b. Partial change: 65% of the weights changed (attention, mlp, norms), in 26 of 40 layers.

Tensors changed: Norms 50/81, MLP 78/120, Attention 104/160

2 derivatives checked

On the Hub · Downloads on Model Pulse