mlabonne/Qwen3-8B-abliterated
LoRA-style vs Qwen/Qwen3-8B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 28 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 84/108, Attention 112/144, lm_head 0/1
7 derivatives checked
- bartowski/mlabonne_Qwen3-8B-abliterated-GGUF: Same weights, converted
- mradermacher/Qwen3-8B-abliterated-GGUF: LoRA-style
- mradermacher/Qwen3-8B-abliterated-i1-GGUF: Full fine-tune
- warshanks/Qwen3-8B-abliterated-AWQ: Quantization
- 123lanhua/mlabonne_Qwen3-8B-abliterated-GGUF: Same weights, converted
- tensorblock/mlabonne_Qwen3-8B-abliterated-GGUF: Quantization
- mclovin/Qwen3-8B-abliterated-Q4_K_M-GGUF: Quantization