Disya/amoral-cogito-Zara-14B
Full fine-tune vs aixonlab/Zara-14b-v1.2. Fine-tune: 98% of the weights changed, across 47 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 95/97, MLP 141/144, Attention 329/336, lm_head 1/1
2 derivatives checked
- mradermacher/amoral-cogito-Zara-14B-i1-GGUF: Quantization
- mradermacher/amoral-cogito-Zara-14B-GGUF: Quantization