Flux output norm projection chain
Flux's final output processing chain combining AdaLayerNormContinuous normalization with output projection. Performs: (1) LayerNorm without affine parameters, (2) SiLU activation on temb followed by linear projection to extract shift/scale modulation parameters, (3) element-wise modulation (shift and scale), (4) large linear projection to output space.
Reported evidence · last observed 2026-08-16. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 16 cases · mean latencyinner_dim = 3072 · output_dim = 64 · modulation_dim = 614416 cases
Not measured on H100 for this workload. Challenges →
145.2µs63% SOL1.00×Reported · license unknown · no source2026-08-16
Measured exactly what you asked. Reported by source; not independently reproduced.
247.5µs62% SOL1.05×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
349.0µs61% SOL1.08×Reported · license unknown · no source2026-07-29
Measured exactly what you asked. Reported by source; not independently reproduced.
451.0µs60% SOL1.13×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
553.6µs59% SOL1.19×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
653.7µs59% SOL1.19×Reported · license unknown · no source2026-07-23
Measured exactly what you asked. Reported by source; not independently reproduced.
754.0µs58% SOL1.20×Reported · license unknown · no source2026-07-15
Measured exactly what you asked. Reported by source; not independently reproduced.
855.5µs57% SOL1.23×Reported · license unknown · no source2026-07-12
Measured exactly what you asked. Reported by source; not independently reproduced.
957.0µs56% SOL1.26×Reported · license unknown · no source2026-08-16
Measured exactly what you asked. Reported by source; not independently reproduced.
1061.6µs54% SOL1.36×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1161.9µs54% SOL1.37×Reported · license unknown · no source2026-08-12
Measured exactly what you asked. Reported by source; not independently reproduced.
1262.6µs54% SOL1.39×Reported · license unknown · no source2026-07-21
Measured exactly what you asked. Reported by source; not independently reproduced.
1362.8µs54% SOL1.39×Reported · license unknown · no source2026-08-08
Measured exactly what you asked. Reported by source; not independently reproduced.
1467.2µs52% SOL1.49×Reported · license unknown · no source2026-07-20
Measured exactly what you asked. Reported by source; not independently reproduced.
1569.3µs51% SOL1.53×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1670.4µs50% SOL1.56×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1770.5µs50% SOL1.56×Reported · license unknown · no source2026-07-24
Measured exactly what you asked. Reported by source; not independently reproduced.
1875.8µs48% SOL1.68×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
19169.8µs28% SOL3.76×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.