Grouped topk MoE routing backward
Backward pass for grouped top-k MoE routing with hierarchical expert selection. Computes gradients through scaling, normalization, gather (scatter), sigmoid, and linear projection.
Reported evidence · last observed 2026-07-10. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 16 cases · mean latencyone = 1 · top_k = 8 · hidden_size = 5120 · n_routed_experts = 16016 cases
Not measured on H100 for this workload. Challenges →
124.0µs80% SOL1.00×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
225.3µs79% SOL1.05×Reported · license unknown · no source2026-08-15
Measured exactly what you asked. Reported by source; not independently reproduced.
325.4µs79% SOL1.06×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
425.7µs78% SOL1.07×Reported · license unknown · no source2026-08-15
Measured exactly what you asked. Reported by source; not independently reproduced.
525.7µs79% SOL1.07×Reported · license unknown · no source2026-07-12
Measured exactly what you asked. Reported by source; not independently reproduced.
625.8µs78% SOL1.08×Reported · license unknown · no source2026-08-15
Measured exactly what you asked. Reported by source; not independently reproduced.
727.0µs77% SOL1.13×Reported · license unknown · no source2026-07-13
Measured exactly what you asked. Reported by source; not independently reproduced.
831.0µs74% SOL1.30×Reported · license unknown · no source2026-07-25
Measured exactly what you asked. Reported by source; not independently reproduced.
931.1µs74% SOL1.30×Reported · license unknown · no source2026-08-14
Measured exactly what you asked. Reported by source; not independently reproduced.
1031.2µs74% SOL1.30×Reported · license unknown · no source2026-08-23
Measured exactly what you asked. Reported by source; not independently reproduced.
1131.3µs74% SOL1.31×Reported · license unknown · no source2026-07-17
Measured exactly what you asked. Reported by source; not independently reproduced.
1231.5µs74% SOL1.31×Reported · license unknown · no source2026-07-11
Measured exactly what you asked. Reported by source; not independently reproduced.
1331.5µs74% SOL1.31×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1431.6µs74% SOL1.32×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.
1532.1µs74% SOL1.34×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1634.9µs72% SOL1.46×Reported · license unknown · no source2026-08-12
Measured exactly what you asked. Reported by source; not independently reproduced.
1735.0µs72% SOL1.46×Reported · license unknown · no source2026-07-28
Measured exactly what you asked. Reported by source; not independently reproduced.
1837.7µs70% SOL1.58×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.
1943.5µs66% SOL1.82×Reported · license unknown · no source2026-08-23
Measured exactly what you asked. Reported by source; not independently reproduced.
2086.4µs49% SOL3.61×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
21533.0µs13% SOL22.25×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.