Skip to content
KernelIndex
Search⌘K

Seqlen-finetuned-reconstructed Hyena gating and output projection

23 eligible runs
moe

Hyena gating mechanism and output projection that combines multiple orders of filtered outputs through cascaded gating with FFT convolution, followed by transpose and linear projection. For order=2: v = dropout(v * x[1]), v = fftconv(v, k[0], bias[0]), y = v * x[0], y = transpose(y), y = out_proj(y).

Fastest reported
44.5µsmean · 82% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-20. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyd_model = 256 · hyena_order = 2 · order_minus_1 = 116 cases

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 16 cases · mean latency · Protocol SOL-ExecBench evaluation stack v1.1 · mean · 16 results · last observed 2026-08-22Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
44.5µs82% SOL
1.00×
Reported · license unknown · no source
2026-08-20

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
2
45.5µs81% SOL
1.02×
Reported · license unknown · no source
2026-07-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
3
46.6µs81% SOL
1.05×
Reported · license unknown · no source
2026-07-25

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
4
60.7µs76% SOL
1.36×
Reported · license unknown · no source
2026-08-05

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
5
61.6µs76% SOL
1.38×
Reported · license unknown · no source
2026-08-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
6
61.9µs76% SOL
1.39×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
7
66.0µs75% SOL
1.48×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
8
73.6µs72% SOL
1.66×
Reported · license unknown · no source
2026-08-22

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
9
75.0µs72% SOL
1.69×
Reported · license unknown · no source
2026-08-16

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
10
75.2µs72% SOL
1.69×
Reported · license unknown · no source
2026-07-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
11
78.0µs71% SOL
1.75×
Reported · license unknown · no source
2026-08-12

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
12
90.9µs68% SOL
2.04×
Reported · license unknown · no source
2026-08-06

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
13
95.9µs67% SOL
2.16×
Reported · license unknown · no source
2026-07-11

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
14
117.4µs62% SOL
2.64×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
15
166.7µs54% SOL
3.75×
Reported · license unknown · no source
2026-07-26

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
16
242.5µs44% SOL
5.45×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
44.5µs
1.00×
Reported
license unknown · no source
unknown
61.2µs
1.38×
Reported
license unknown · no source
unknown
61.6µs
1.38×
Reported
license unknown · no source
unknown
66.0µs
1.48×
Reported
license unknown · no source
unknown
73.6µs
1.66×
Reported
license unknown · no source
unknown
95.9µs
2.16×
Reported
license unknown · no source
unknown
238.5µs
5.36×
Reported
license unknown · no source
unknown
61.9µs
1.39×
Reported
license unknown · no source
Show all 23 implementations ›
unknown
66.0µs
1.48×
Reported
license unknown · no source
unknown
75.0µs
1.69×
Reported
license unknown · no source
unknown
75.2µs
1.69×
Reported
license unknown · no source
unknown
78.0µs
1.75×
Reported
license unknown · no source
unknown
81.9µs
1.84×
Reported
license unknown · no source
unknown
90.9µs
2.04×
Reported
license unknown · no source
unknown
117.3µs
2.64×
Reported
license unknown · no source
unknown
117.4µs
2.64×
Reported
license unknown · no source
unknown
166.7µs
3.75×
Reported
license unknown · no source
unknown
180.3µs
4.06×
Reported
license unknown · no source
unknown
242.5µs
5.45×
Reported
license unknown · no source
unknown
243.8µs
5.48×
Reported
license unknown · no source
unknown
45.5µs
1.02×
Reported
license unknown · no source
unknown
46.6µs
1.05×
Reported
license unknown · no source
unknown
60.7µs
1.36×
Reported
license unknown · no source

Semantics

Inputs and outputs
vfp32 [batch_size, d_model, seq_len]
x0fp32 [batch_size, d_model, seq_len]
x1fp32 [batch_size, d_model, seq_len]
kfp32 [order_minus_1, d_model, seq_len]
biasfp32 [order_minus_1, d_model]
out_proj_weightfp32 [d_model, d_model]
out_proj_biasfp32 [d_model]
outputfp32 [batch_size, seq_len, d_model]
Axes and behavior
d_modelconstant = 256
seq_lenvariable
batch_sizevariable
hyena_orderconstant = 2
fft_size_exprfft_size_expr = seq_len * 2
order_minus_1constant = 1
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias052_seqlen-finetuned-reconstructed_hyena_gating_and_output_projectionmodelhyena-small-32ksha256d4d2d037ed8b…
Sources: NVIDIA SOL-ExecBench (2026-08-22)last observed 2026-08-22How records are decidedJSON