mergekit-community/Irix-12B_Slush
Partial fine-tune vs DreadPoor/Irix-12B-Model_Stock. Partial change: 85% of the weights changed (attention, mlp, norms), in 38 of 40 layers.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 63/81, MLP 114/120, Attention 152/160
2 derivatives checked
- mradermacher/Irix-12B_Slush-i1-GGUF: Quantization
- mradermacher/Irix-12B_Slush-GGUF: Quantization