Skip to content
KernelIndex
Search⌘K
Sign in

GIT vision encoder layer

26 eligible runs
vision

Complete vision encoder layer combining multi-head self-attention with pre-LayerNorm, MLP with GELU activation, and residual connections. Performs: pre-norm -> attention -> residual -> pre-norm -> MLP -> residual.

Fastest reported
168.8µsmean · 61% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-14. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhead_dim = 64 · num_heads = 12 · hidden_size = 768 · intermediate_size = 307216 cases
seq_len
batch_size
dtype
1971fp321972fp321978fp3219716fp3219732fp3219764fp322112fp3244916fp325124fp3254132fp3261316fp329198fp3212498fp32142332fp3220481fp3240962fp32
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 1168.8 µsRIAC as well19NVIDIA B200 · env 2173.2 µsRecursive7

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 · 19 results · last observed 2026-08-15Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
RIAC as well
168.8µs61% SOL
1.00×
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 →
2
Recursive
186.6µs58% SOL
1.11×
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
Amir M. Mir | SF Tensor
189.1µs57% SOL
1.12×
Reported · license unknown · no source
2026-08-01

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

no sourcelicense unknownno install recipeRun detail →
4
Amir M. Mir | SF Tensor
193.7µs56% SOL
1.15×
Reported · license unknown · no source
2026-08-01

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

no sourcelicense unknownno install recipeRun detail →
5
doubleAI
197.8µs56% SOL
1.17×
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 →
6
Geometric
199.1µs56% SOL
1.18×
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 →
7
Proud Wren @ NVIDIA
201.2µs56% SOL
1.19×
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 →
8
Amir M. Mir | SF Tensor
207.4µs55% SOL
1.23×
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 →
9
ac4k
215.5µs53% SOL
1.28×
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 →
10
ac4k
218.1µs53% SOL
1.29×
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 →
11
Electric Cheetah
218.6µs53% SOL
1.30×
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 →
12
Rapid Tiger
218.7µs53% SOL
1.30×
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
Gleaming Coyote
218.8µs53% SOL
1.30×
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 →
14
Hyra
220.9µs53% SOL
1.31×
Reported · license unknown · no source
2026-07-19

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

no sourcelicense unknownno install recipeRun detail →
15
Cursor
246.2µs49% SOL
1.46×
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
Geometric
549.7µs28% SOL
3.26×
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 →
17
Geometric
1.36ms14% SOL
8.03×
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 →
18
Geometric
1.72ms11% SOL
10.19×
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 →
19
Geometric
1.86ms10% SOL
11.04×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
RIAC as well
unknown
168.8µs
1.00×
Reported
license unknown · no source
Recursive
unknown
186.6µs
1.11×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
189.1µs
1.12×
Reported
license unknown · no source
Recursive
unknown
173.2µs
1.03×
Reported
license unknown · no source
doubleAI
unknown
193.0µs
1.14×
Reported
license unknown · no source
Geometric
unknown
199.1µs
1.18×
Reported
license unknown · no source
doubleAI
unknown
197.8µs
1.17×
Reported
license unknown · no source
Proud Wren @ NVIDIA
unknown
201.2µs
1.19×
Reported
license unknown · no source
Show all 26 implementations ›
ac4k
unknown
215.5µs
1.28×
Reported
license unknown · no source
Electric Cheetah
unknown
218.6µs
1.30×
Reported
license unknown · no source
Rapid Tiger
unknown
218.7µs
1.30×
Reported
license unknown · no source
Gleaming Coyote
unknown
218.8µs
1.30×
Reported
license unknown · no source
Hyra
unknown
220.9µs
1.31×
Reported
license unknown · no source
Cursor
unknown
232.3µs
1.38×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
193.7µs
1.15×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
202.9µs
1.20×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
207.4µs
1.23×
Reported
license unknown · no source
ac4k
unknown
218.1µs
1.29×
Reported
license unknown · no source
Cursor
unknown
246.2µs
1.46×
Reported
license unknown · no source
doubleAI
unknown
256.5µs
1.52×
Reported
license unknown · no source
Geometric
unknown
549.7µs
3.26×
Reported
license unknown · no source
Geometric
unknown
1.36ms
8.03×
Reported
license unknown · no source
Geometric
unknown
1.72ms
10.19×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
1.96ms
11.60×
Reported
license unknown · no source
Geometric
unknown
1.86ms
11.04×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
2.17ms
12.88×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesfp32 [batch_size, seq_len, hidden_size]
layer_norm1_weightfp32 [hidden_size]
layer_norm1_biasfp32 [hidden_size]
q_proj_weightfp32 [hidden_size, hidden_size]
q_proj_biasfp32 [hidden_size]
k_proj_weightfp32 [hidden_size, hidden_size]
k_proj_biasfp32 [hidden_size]
v_proj_weightfp32 [hidden_size, hidden_size]
v_proj_biasfp32 [hidden_size]
out_proj_weightfp32 [hidden_size, hidden_size]
out_proj_biasfp32 [hidden_size]
layer_norm2_weightfp32 [hidden_size]
layer_norm2_biasfp32 [hidden_size]
fc1_weightfp32 [intermediate_size, hidden_size]
fc1_biasfp32 [intermediate_size]
fc2_weightfp32 [hidden_size, intermediate_size]
fc2_biasfp32 [hidden_size]
layer_norm_epsfp32 scalar
outputfp32 [batch_size, seq_len, hidden_size]
Axes and behavior
seq_lenvariable
head_dimhead_dim = hidden_size // num_heads
num_headsconstant = 12
batch_sizevariable
hidden_sizeconstant = 768
intermediate_sizeconstant = 3072
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias079_git_vision_encoder_layermodelmaskgit-imagenet-soft-code-34ksha256d8aead85f57f…
Sources: NVIDIA SOL-ExecBench (2026-08-15)last observed 2026-08-15How records are decidedJSON