stepenZEN/DeepSeek-R1-Distill-Llama-8B-Abliterated
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Llama-8B. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 32 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Attention 32/128, MLP 32/96, Norms 0/65, lm_head 0/1
8 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Llama-8B-Abliterated-GGUF: Same weights, converted
- mradermacher/DeepSeek-R1-Distill-Llama-8B-Abliterated-i1-GGUF: Quantization
- ufydy937/DeepSeek-R1-Distill-Llama-8B-Abliterated-GGUF: Same weights, converted
- data-pioneer-0826/DeepSeek-R1-Distill-Llama-8B-Abliterated-GGUF: Same weights, converted
- matrixportalx/DeepSeek-R1-Distill-Llama-8B-Abliterated-GGUF: Quantization
- fairy322/DeepSeek-R1-Distill-Llama-8B-i1-GGUF: Quantization
- ibrahimkettaneh/DeepSeek-R1-Distill-Llama-8B-Abliterated-Q4_K_M-GGUF: Quantization
- Yuma42/Llama3.1-CrimeSolver-8B: Full fine-tune