Skip to content
KernelIndex
Search⌘K

Vision encoder layer with gated residuals

27 eligible runs
vision

Complete vision encoder layer with self-attention, MLP, layer norms, and tanh-gated residual connections. This is the dominant computational pattern in the vision encoder, appearing in 8 global vision layers with gating. The gated residuals use tanh activation on learnable gate parameters to modulate the contribution of attention and MLP outputs.

Fastest reported
175.7µsmean · 66% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-16. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhead_dim = 80 · num_heads = 16 · hidden_size = 1280 · intermediate_size = 512016 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 · 20 results · last observed 2026-08-17Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
175.7µs66% SOL
1.00×
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 →
2
190.5µs63% SOL
1.08×
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 →
3
216.8µs59% SOL
1.23×
Reported · license unknown · no source
2026-07-21

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

no sourcelicense unknownno install recipeRun detail →
4
219.2µs59% SOL
1.25×
Reported · license unknown · no source
2026-07-21

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

no sourcelicense unknownno install recipeRun detail →
5
223.5µs58% SOL
1.27×
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 →
6
226.8µs58% SOL
1.29×
Reported · license unknown · no source
2026-07-17

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

no sourcelicense unknownno install recipeRun detail →
7
227.4µs57% SOL
1.29×
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
228.7µs58% SOL
1.30×
Reported · license unknown · no source
2026-08-17

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

no sourcelicense unknownno install recipeRun detail →
9
253.2µs54% SOL
1.44×
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 →
10
277.2µs52% SOL
1.58×
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 →
11
279.1µs50% 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 →
12
334.2µs44% SOL
1.90×
Reported · license unknown · no source
2026-08-15

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

no sourcelicense unknownno install recipeRun detail →
13
360.7µs43% SOL
2.05×
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 →
14
391.4µs40% SOL
2.23×
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 →
15
495.1µs34% SOL
2.82×
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 →
16
521.6µs33% SOL
2.97×
Reported · license unknown · no source
2026-07-23

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

no sourcelicense unknownno install recipeRun detail →
17
934.9µs21% SOL
5.32×
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 →
18
1.58ms15% SOL
8.99×
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 →
19
1.61ms16% SOL
9.18×
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 →
20
1.70ms15% SOL
9.68×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
175.7µs
1.00×
Reported
license unknown · no source
unknown
185.8µs
1.06×
Reported
license unknown · no source
unknown
219.2µs
1.25×
Reported
license unknown · no source
unknown
228.7µs
1.30×
Reported
license unknown · no source
unknown
226.8µs
1.29×
Reported
license unknown · no source
unknown
253.2µs
1.44×
Reported
license unknown · no source
unknown
268.3µs
1.53×
Reported
license unknown · no source
unknown
455.5µs
2.59×
Reported
license unknown · no source
Show all 27 implementations ›
unknown
223.5µs
1.27×
Reported
license unknown · no source
unknown
227.4µs
1.29×
Reported
license unknown · no source
unknown
269.8µs
1.54×
Reported
license unknown · no source
unknown
277.2µs
1.58×
Reported
license unknown · no source
unknown
279.0µs
1.59×
Reported
license unknown · no source
unknown
279.1µs
1.59×
Reported
license unknown · no source
unknown
334.2µs
1.90×
Reported
license unknown · no source
unknown
347.8µs
1.98×
Reported
license unknown · no source
unknown
360.7µs
2.05×
Reported
license unknown · no source
unknown
391.4µs
2.23×
Reported
license unknown · no source
unknown
495.1µs
2.82×
Reported
license unknown · no source
unknown
521.6µs
2.97×
Reported
license unknown · no source
unknown
934.9µs
5.32×
Reported
license unknown · no source
unknown
1.61ms
9.18×
Reported
license unknown · no source
unknown
1.58ms
8.99×
Reported
license unknown · no source
unknown
1.70ms
9.68×
Reported
license unknown · no source
unknown
190.5µs
1.08×
Reported
license unknown · no source
unknown
213.3µs
1.21×
Reported
license unknown · no source
unknown
216.8µs
1.23×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statebf16 [batch_size, seq_len, hidden_size]
input_layernorm_weightbf16 [hidden_size]
input_layernorm_biasbf16 [hidden_size]
q_proj_weightbf16 [hidden_size, hidden_size]
k_proj_weightbf16 [hidden_size, hidden_size]
v_proj_weightbf16 [hidden_size, hidden_size]
o_proj_weightbf16 [hidden_size, hidden_size]
post_attention_layernorm_weightbf16 [hidden_size]
post_attention_layernorm_biasbf16 [hidden_size]
fc1_weightbf16 [intermediate_size, hidden_size]
fc1_biasbf16 [intermediate_size]
fc2_weightbf16 [hidden_size, intermediate_size]
fc2_biasbf16 [hidden_size]
gate_attnbf16 [1]
gate_ffnbf16 [1]
norm_epsfp32 scalar
outputbf16 [batch_size, seq_len, hidden_size]
Axes and behavior
seq_lenvariable
head_dimhead_dim = hidden_size // num_heads
num_headsconstant = 16
batch_sizevariable
hidden_sizeconstant = 1280
intermediate_sizeconstant = 5120
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias054_vision_encoder_layer_with_gated_residualsmodelllama-3-2-11b-visionsha256127ebd3cfe89…
Sources: NVIDIA SOL-ExecBench (2026-08-17)last observed 2026-08-17How records are decidedJSON