jsfs11/MixtureofMerges-MoE-4x7bRP-v11
MoE built from dense models vs BioMistral/BioMistral-DARE-NS. jsfs11/MixtureofMerges-MoE-4x7bRP-v11 is a mixture of 4 experts assembled from dense models (mergekit-moe style): expert 2 is an exact copy of BioMistral/BioMistral-DARE-NS's MLP; the other experts come from other models. Shared tensors (attention, norms, embeddings): 0 of 193 identical.
2 derivatives checked
- mradermacher/MixtureofMerges-MoE-4x7bRP-v11-i1-GGUF: Quantization
- mradermacher/MixtureofMerges-MoE-4x7bRP-v11-GGUF: Quantization