kakaocorp/kanana-2-30b-a3b-instruct-2601

Full fine-tune vs kakaocorp/kanana-2-30b-a3b-mid-2601. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/1, Norms 145/145, MLP 2614/2661, Attention 192/192, lm_head 1/1

3 derivatives checked

• deul/Kanana-2-30b-a3b-instruct-2601-GGUF: Quantization
• ohora23/Kanana-2-30B-A3B-Instruct-GGUF: Quantization
• NotoriousH2/kanana-2-30b-a3b-instruct-2601-awq-w4a16: Quantization

On the Hub · Downloads on Model Pulse