DavidAU/Qwen3-MOE-4x1.7B-6.8B-Lucys-Song-Uncensored
MoE built from dense models vs Goekdeniz-Guelmez/Josiefied-Qwen3-1.7B-abliterated-v1. DavidAU/Qwen3-MOE-4x1.7B-6.8B-Lucys-Song-Uncensored is a mixture of 4 experts assembled from dense models (mergekit-moe style): expert 3 is an exact copy of Goekdeniz-Guelmez/Josiefied-Qwen3-1.7B-abliterated-v1's MLP; the other experts come from other models. Shared tensors (attention, norms, embeddings): 16 of 226 identical.
2 derivatives checked
- mradermacher/Qwen3-MOE-4x1.7B-6.8B-Lucys-Song-Uncensored-i1-GGUF: Quantization
- mradermacher/Qwen3-MOE-4x1.7B-6.8B-Lucys-Song-Uncensored-GGUF: Same weights, converted