huihui-ai/Huihui-Qwen3.5-4B-abliterated
LoRA-style vs Qwen/Qwen3.5-4B. 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, out_proj, q_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, MLP 64/196, Attention 56/320, Norms 0/203
10 derivatives checked
- mradermacher/Huihui-Qwen3.5-4B-abliterated-GGUF: Same weights, converted
- mradermacher/Huihui-Qwen3.5-4B-abliterated-i1-GGUF: Quantization
- interimlabs/InterimLabs-Huihui-Qwen3.5-4B-abliterated-GGUF-Q4_K_M: Quantization
- TheWirelessPhoenix/Qwen3.5-4B-abliterated_4bit: Quantization
- interimlabs/InterimLabs-Huihui-Qwen3.5-4B-abliterated-MLX-4bit: Quantization
- Ayaz672i/Huihui-Qwen3.5-4B-Abliterated-PMRA-GGUF: Quantization
- Asystemoffields/Huihui-Qwen3.5-4B-Abliterated-PMRA-GGUF: Quantization
- Etherealty/Huihui-Qwen3.5-4B-Abliterated-PMRA-GGUF: Quantization
- nbeerbower/Huihui-Qwen3.5-4B-abliterated-Athanorlite-ORPO: LoRA-style
- elyssiaelysium/elyssia: LoRA-style
2 other repos hold byte-for-byte identical weight files.