harkov000/hermes-stheno-8B-v0.1

Partial fine-tune vs NousResearch/Hermes-3-Llama-3.1-8B. Partial change: 81% of the weights changed (attention, mlp, norms), in 30 of 32 layers.

Tensors changed: Embeddings 0/1, Norms 36/65, MLP 90/96, Attention 120/128, lm_head 0/1

2 derivatives checked