nlpai-lab/KULLM3

Full fine-tune vs upstage/SOLAR-10.7B-v1.0. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/1, Norms 97/97, MLP 144/144, Attention 192/192, lm_head 1/1

2 derivatives checked

• featherless-ai-quants/nlpai-lab-KULLM3-GGUF: Quantization
• tensorblock/KULLM3-GGUF: Quantization

On the Hub · Downloads on Model Pulse