mergekit-community/mergekit-slerp-xlkdlta
Full fine-tune vs Equall/Saul-7B-Base. Fine-tune: 97% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 16/65, MLP 93/96, Attention 124/128, lm_head 1/1
1 derivatives checked
- MaziyarPanahi/mergekit-slerp-xlkdlta-GGUF: Quantization
37 other repos hold byte-for-byte identical weight files.