tbilisi-ai-lab/kona2-12B-Base
Partial fine-tune vs mistralai/Mistral-Nemo-Base-2407. Partial change: 72% of the weights changed (attention, mlp, norms), in 40 of 40 layers.
Tensors changed: Norms 1/81, MLP 88/120, Attention 127/160
3 derivatives checked
- mradermacher/kona2-12B-Base-i1-GGUF: Quantization
- mradermacher/kona2-12B-Base-GGUF: Quantization
- tbilisi-ai-lab/kona2-12B-Instruct: Full fine-tune