nomic_bert_moe_gguf¶
Model type |
|
Class |
|
Task |
|
Source |
|
Description¶
NomicBERT-MoE encoder with periodic routed, non-gated GELU experts.
Routed layers are selected by layer % moe_every_n_layers == 1.
Router probabilities are softmaxed over every expert before top-k
selection and are not renormalized after selection.
Usage¶
mobius build --model <MODEL_ID> output_dir/
from mobius import build
model = build("<MODEL_ID>")