mergekit-community/Deepseek-Distill-NSFW-visible-w-NSFW-FFS
Full fine-tune vs mergekit-community/Deepseek-R1-Distill-NSFW-RPv1. Fine-tune: 91% of the weights changed, across 31 of 32 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 63/65, MLP 93/96, Attention 124/128
2 derivatives checked
- mradermacher/Deepseek-Distill-NSFW-visible-w-NSFW-FFS-i1-GGUF: Quantization
- mradermacher/Deepseek-Distill-NSFW-visible-w-NSFW-FFS-GGUF: Same weights, converted