rAIfle/WAIDWML-Phi4-8x14B-bf16
MoE built from dense models vs unsloth/phi-4. rAIfle/WAIDWML-Phi4-8x14B-bf16 is a mixture of 8 experts with the same shape as unsloth/phi-4'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): 83 of 243 identical.
1 derivatives checked
- mradermacher/WAIDWML-Phi4-8x14B-bf16-i1-GGUF: Quantization