Sakalti/SakaMoe-3x14B-Instruct
MoE built from dense models vs Sakalti/Saka-14B. Sakalti/SakaMoe-3x14B-Instruct is a mixture of 3 experts assembled from dense models (mergekit-moe style): experts 0, 2 are an exact copy of Sakalti/Saka-14B's MLP; the other experts come from other models. Shared tensors (attention, norms, embeddings): 435 of 435 identical.
1 derivatives checked
- mradermacher/SakaMoe-3x14B-Instruct-GGUF: Quantization