SteelStorage/phi-2-DLEC

Partial fine-tune vs abacaj/phi-2-super. Partial change: 68% of the weights changed (attention, mlp, norms), in 24 of 32 layers.

Tensors changed: Embeddings 0/1, Norms 48/66, MLP 96/128, Attention 192/256, lm_head 0/2

1 derivatives checked

On the Hub · Downloads on Model Pulse