mergekit-community/mergekit-passthrough-dmirwnd
Full fine-tune (fewer layers) vs NousResearch/Hermes-2-Pro-Llama-3-8B. NousResearch/Hermes-2-Pro-Llama-3-8B matches part of mergekit-community/mergekit-passthrough-dmirwnd (290 of its 498 tensors; the rest of mergekit-community/mergekit-passthrough-dmirwnd has no counterpart). On the shared part: Fine-tune: 93% of the weights changed, across 32 of 32 layers.
Tensors changed: Norms 64/65, MLP 96/96, Attention 128/128, lm_head 0/1
1 derivatives checked
- MaziyarPanahi/mergekit-passthrough-dmirwnd-GGUF: Quantization