jsfs11/QWQ-EVA-Preview-SLERP-32b

Full fine-tune vs ParasiticRogue/EVA-Instruct-32B-v2. Fine-tune: 92% of the weights changed, across 62 of 64 layers.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 123/129, MLP 186/192, Attention 434/448

2 derivatives checked

On the Hub · Downloads on Model Pulse