Skip to content
KernelIndex
Search⌘K
Sign in

Vision MLP GELU backward

32 eligible runs
mlp

Backward pass for Vision MLP with GELU activation. Computes gradients for input hidden state, fc1 weight/bias, and fc2 weight/bias through the chain: fc2 -> GELU -> fc1.

Fastest reported
71.9µsmean · 70% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-24. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhidden_size = 1280 · intermediate_size = 358416 cases
seq_len
dtype
64fp32128fp32131fp32211fp32256fp32293fp32449fp32512fp32768fp32773fp32997fp321163fp321423fp321973fp322048fp324096fp32
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 171.9 µsRIAC as well22NVIDIA B200 · env 278.5 µsdoubleAI10

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 · 22 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
RIAC as well
71.9µs70% SOL
1.00×
Reported · license unknown · no source
2026-08-24

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

no sourcelicense unknownno install recipeRun detail →
2
Hyra
81.6µs67% SOL
1.13×
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
doubleAI
82.0µs66% SOL
1.14×
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 →
4
Databricks
87.2µs65% SOL
1.21×
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 →
5
TH3CHARLie | TileGym
87.2µs65% SOL
1.21×
Reported · license unknown · no source
2026-07-20

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

no sourcelicense unknownno install recipeRun detail →
6
Amir M. Mir | SF Tensor
89.9µs64% SOL
1.25×
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
Argus
90.8µs64% SOL
1.26×
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
ac4k
95.1µs62% SOL
1.32×
Reported · license unknown · no source
2026-07-12

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

no sourcelicense unknownno install recipeRun detail →
9
Geometric
103.0µs60% SOL
1.43×
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 →
10
ac4k
105.8µs59% SOL
1.47×
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 →
11
Amir M. Mir | SF Tensor
108.9µs59% SOL
1.51×
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 →
12
Recursive
114.0µs57% SOL
1.59×
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 →
13
Geometric
122.2µs56% SOL
1.70×
Reported · license unknown · no source
2026-08-24

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

no sourcelicense unknownno install recipeRun detail →
14
Geometric
126.6µs54% SOL
1.76×
Reported · license unknown · no source
2026-08-07

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

no sourcelicense unknownno install recipeRun detail →
15
Geometric
132.4µs53% SOL
1.84×
Reported · license unknown · no source
2026-07-24

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

no sourcelicense unknownno install recipeRun detail →
16
Geometric
138.0µs52% SOL
1.92×
Reported · license unknown · no source
2026-08-10

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

no sourcelicense unknownno install recipeRun detail →
17
Geometric
144.5µs51% SOL
2.01×
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 →
18
Geometric
144.7µs51% SOL
2.01×
Reported · license unknown · no source
2026-08-11

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

no sourcelicense unknownno install recipeRun detail →
19
Cursor
181.0µs45% SOL
2.52×
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 →
20
ac4k
191.4µs43% SOL
2.66×
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 →
21
Geometric
329.9µs30% SOL
4.59×
Reported · license unknown · no source
2026-08-13

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

no sourcelicense unknownno install recipeRun detail →
22
Runic Kingfisher
1.24ms10% SOL
17.20×
Reported · license unknown · no source
2026-08-02

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

no sourcelicense unknownno install recipeRun detail →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
RIAC as well
unknown
71.9µs
1.00×
Reported
license unknown · no source
doubleAI
unknown
78.5µs
1.09×
Reported
license unknown · no source
Databricks
unknown
84.4µs
1.17×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
89.9µs
1.25×
Reported
license unknown · no source
Argus
unknown
89.1µs
1.24×
Reported
license unknown · no source
ac4k
unknown
89.9µs
1.25×
Reported
license unknown · no source
Geometric
unknown
103.0µs
1.43×
Reported
license unknown · no source
Recursive
unknown
111.0µs
1.55×
Reported
license unknown · no source
Show all 32 implementations ›
Cursor
unknown
172.4µs
2.40×
Reported
license unknown · no source
Igor Bragaia
unknown
259.2µs
3.61×
Reported
license unknown · no source
Runic Kingfisher
unknown
1.24ms
17.20×
Reported
license unknown · no source
Databricks
unknown
87.2µs
1.21×
Reported
license unknown · no source
Argus
unknown
90.8µs
1.26×
Reported
license unknown · no source
ac4k
unknown
95.1µs
1.32×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
105.0µs
1.46×
Reported
license unknown · no source
ac4k
unknown
105.8µs
1.47×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
108.9µs
1.51×
Reported
license unknown · no source
Recursive
unknown
114.0µs
1.59×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
118.1µs
1.64×
Reported
license unknown · no source
Geometric
unknown
122.2µs
1.70×
Reported
license unknown · no source
Geometric
unknown
126.6µs
1.76×
Reported
license unknown · no source
Geometric
unknown
132.4µs
1.84×
Reported
license unknown · no source
Geometric
unknown
138.0µs
1.92×
Reported
license unknown · no source
Geometric
unknown
144.5µs
2.01×
Reported
license unknown · no source
Geometric
unknown
144.7µs
2.01×
Reported
license unknown · no source
Cursor
unknown
181.0µs
2.52×
Reported
license unknown · no source
ac4k
unknown
191.4µs
2.66×
Reported
license unknown · no source
Geometric
unknown
329.9µs
4.59×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
657.7µs
9.15×
Reported
license unknown · no source
Hyra
unknown
81.6µs
1.13×
Reported
license unknown · no source
doubleAI
unknown
82.0µs
1.14×
Reported
license unknown · no source
TH3CHARLie | TileGym
unknown
87.2µs
1.21×
Reported
license unknown · no source

Semantics

Inputs and outputs
grad_outputfp32 [seq_len, hidden_size]
hidden_statefp32 [seq_len, hidden_size]
fc1_weightfp32 [intermediate_size, hidden_size]
fc1_biasfp32 [intermediate_size]
fc2_weightfp32 [hidden_size, intermediate_size]
fc2_biasfp32 [hidden_size]
fc1_outputfp32 [seq_len, intermediate_size]
gelu_outputfp32 [seq_len, intermediate_size]
grad_hidden_statefp32 [seq_len, hidden_size]
grad_fc1_weightfp32 [intermediate_size, hidden_size]
grad_fc1_biasfp32 [intermediate_size]
grad_fc2_weightfp32 [hidden_size, intermediate_size]
grad_fc2_biasfp32 [hidden_size]
Axes and behavior
seq_lenvariable
hidden_sizeconstant = 1280
intermediate_sizeconstant = 3584
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias022_vision_mlp_gelu_backwardmodelqwen3-vl-8b-instructsha2569e287b1df00c…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON