Disya/Xgen-9B-Merge
Partial fine-tune vs Delta-Vector/Austral-Xgen-9B-Winton. Partial change: 58% of the weights changed (attention, lm_head, mlp, norms), in 44 of 45 layers.
Tensors changed: Embeddings 0/1, Norms 42/91, MLP 74/135, Attention 136/180, lm_head 1/1
1 derivatives checked
- mradermacher/Xgen-9B-Merge-GGUF: Same weights, converted