Trace the router
Record selected routed experts, weights, shared experts, layer, generation step, and model provenance.
Project journal · Phase 01 · In progress
MoEscope follows Mixture-of-Experts inference from a token’s routing decision to the GPU workload it creates—then makes the evidence inspectable.
Record selected routed experts, weights, shared experts, layer, generation step, and model provenance.
Turn assignments into expert histograms, grouped matrix shapes, tile counts, padding, and theoretical waves.
Replay the same workload against a correct reference and measured GPU backends without hiding losses.