jsfs11/MixtureofMerges-MoE-4x7b-v4
MoE built from dense models vs Kukedlc/NeuTrixOmniBe-7B-model-remix. jsfs11/MixtureofMerges-MoE-4x7b-v4 is a mixture of 4 experts assembled from dense models (mergekit-moe style): expert 1 is an exact copy of Kukedlc/NeuTrixOmniBe-7B-model-remix's MLP; the other experts come from other models. Shared tensors (attention, norms, embeddings): 195 of 195 identical.
2 derivatives checked
- mradermacher/MixtureofMerges-MoE-4x7b-v4-i1-GGUF: Quantization
- mradermacher/MixtureofMerges-MoE-4x7b-v4-GGUF: Quantization