Davidsv/SUONG-3
Full fine-tune vs OpenPipe/mistral-ft-optimized-1218. Fine-tune: 97% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 22/65, MLP 93/96, Attention 124/128, lm_head 1/1
1 derivatives checked
- mradermacher/SUONG-3-GGUF: Same weights, converted
22 other repos hold byte-for-byte identical weight files.