jeiku/zenithquantum
Partial fine-tune vs FourOhFour/QuantuMinx_4B. Partial change: 89% of the weights changed (attention, lm_head, mlp, norms), in 32 of 32 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 21/65, MLP 93/96, Attention 124/128
1 derivatives checked
- mradermacher/zenithquantum-GGUF: Same weights, converted