kakaocorp/kanana-2-3b-instruct
Full fine-tune vs kakaocorp/kanana-2-3b-base. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 129/129, MLP 96/96, Attention 128/128
3 derivatives checked
- mradermacher/kanana-2-3b-instruct-GGUF: Same weights, converted
- mradermacher/kanana-2-3b-instruct-i1-GGUF: Quantization
- mlx-community/kanana-2-3b-instruct-4bit: Quantization