mergekit-community/L3.1-15B-EtherealMaid-t0.0001-alpha
Partial fine-tune (fewer layers) vs Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2. Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2 matches part of mergekit-community/L3.1-15B-EtherealMaid-t0.0001-alpha (291 of its 579 tensors; the rest of mergekit-community/L3.1-15B-EtherealMaid-t0.0001-alpha has no counterpart). On the shared part: Partial change: 71% of the weights changed (attention, mlp, norms), in 26 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 52/65, MLP 78/96, Attention 104/128, lm_head 0/1
2 derivatives checked
- mradermacher/L3.1-15B-EtherealMaid-t0.0001-alpha-i1-GGUF: Quantization
- mradermacher/L3.1-15B-EtherealMaid-t0.0001-alpha-GGUF: Quantization