Jebadiah/dolphin-hermes-stone-l3-8b-1m
Partial fine-tune vs NousResearch/Hermes-2-Pro-Llama-3-8B. Partial change: 89% of the weights changed (attention, mlp, norms), in 32 of 32 layers.
Tensors changed: Norms 12/65, MLP 84/96, Attention 113/128
1 derivatives checked
- mradermacher/dolphin-hermes-stone-l3-8b-1m-GGUF: Same weights, converted