Nexesenex/Llama_3.2_3b_Kermes_0.20

Full fine-tune vs MaziyarPanahi/calme-3.3-llamaloi-3b. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: MLP 84/84, Embeddings 1/1, Norms 57/57, Attention 112/112

2 derivatives checked

• mradermacher/Llama_3.2_3b_Kermes_0.2_bf16-GGUF: Same weights, converted
• Nexesenex/Llama_3.2_3b_Kermes_v2.1: Full fine-tune

On the Hub · Downloads on Model Pulse