Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1
Abliterated vs Qwen/Qwen2.5-Coder-7B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 28 of 28 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Norms 0/57, MLP 28/84, Attention 28/196, lm_head 0/1
6 derivatives checked
- mradermacher/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-i1-GGUF: Quantization
- Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-gguf: Quantization
- mradermacher/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-GGUF: Same weights, converted
- mlx-community/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-4bit: Quantization
- mlx-community/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-8bit: Quantization
- mlx-community/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1: Same weights, converted