bunnycore/Phi-2-Hyper
Full fine-tune vs Artples/L-MChat-Small. Fine-tune: 97% of the weights changed, across 32 of 32 layers.
Tensors changed: lm_head 2/2, Embeddings 1/1, Norms 53/66, MLP 124/128, Attention 248/256
1 derivatives checked
- mradermacher/Phi-2-Hyper-GGUF: Same weights, converted