gghfez/WizardLM-2-22b-RP
MoE built from dense models vs alpindale/WizardLM-2-8x22B. alpindale/WizardLM-2-8x22B is a mixture of 8 experts with the same shape as gghfez/WizardLM-2-22b-RP's MLP, but none of them is a copy of it: they come from other models or were trained further. Shared tensors (attention, norms, embeddings): 103 of 337 identical.
3 derivatives checked
- mradermacher/WizardLM-2-22b-RP-i1-GGUF: Quantization
- gghfez/WizardLM-2-22B-RP-control-vectors: Not comparable yet
- mradermacher/WizardLM-2-22b-RP-GGUF: Quantization