jsfs11/MixtureofMerges-MoE-4x7b-v3

MoE built from dense models vs jsfs11/RandomMergeNoNormWEIGHTED-7B-DARETIES. jsfs11/MixtureofMerges-MoE-4x7b-v3 is a mixture of 4 experts assembled from dense models (mergekit-moe style): expert 0 is an exact copy of jsfs11/RandomMergeNoNormWEIGHTED-7B-DARETIES's MLP; the other experts come from other models. Shared tensors (attention, norms, embeddings): 0 of 195 identical.

2 derivatives checked

On the Hub · Downloads on Model Pulse