Skip to content
KernelIndex
Search⌘K

Video spatial attention with RoPE 3d

34 eligible runs
rope

Video Spatial Attention with 3D Rotary Position Embedding. Applies multi-head self-attention to video tokens with 3D RoPE encoding spatial (height, width) and temporal (frame) positions. Critical for text-to-video generation where spatial coherence and temporal awareness are essential.

Fastest reported
132.6µsmean · 89% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-19. 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 · qkv_hidden = 3072 · hidden_size = 1024 · half_head_dim = 32 · num_attention_heads = 1616 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 · 19 results · last observed 2026-08-23Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
132.6µs89% SOL
1.00×
Reported · license unknown · no source
2026-08-19

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

no sourcelicense unknownno install recipeRun detail →
2
142.1µs88% SOL
1.07×
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
149.3µs88% SOL
1.13×
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
149.7µs88% SOL
1.13×
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
154.8µs87% SOL
1.17×
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 →
6
155.8µs87% 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 →
7
166.3µs86% SOL
1.25×
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
169.5µs86% SOL
1.28×
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
171.1µs86% SOL
1.29×
Reported · license unknown · no source
2026-08-18

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

no sourcelicense unknownno install recipeRun detail →
10
172.1µs86% SOL
1.30×
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 →
11
177.1µs85% SOL
1.34×
Reported · license unknown · no source
2026-08-23

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

no sourcelicense unknownno install recipeRun detail →
12
180.4µs85% SOL
1.36×
Reported · license unknown · no source
2026-08-18

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

no sourcelicense unknownno install recipeRun detail →
13
181.6µs85% SOL
1.37×
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 →
14
183.1µs85% SOL
1.38×
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 →
15
195.5µs83% SOL
1.47×
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 →
16
196.9µs83% 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 →
17
204.3µs82% SOL
1.54×
Reported · license unknown · no source
2026-07-27

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

no sourcelicense unknownno install recipeRun detail →
18
732.1µs56% SOL
5.52×
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 →
19
1.68ms35% SOL
12.69×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
132.6µs
1.00×
Reported
license unknown · no source
unknown
133.8µs
1.01×
Reported
license unknown · no source
unknown
149.8µs
1.13×
Reported
license unknown · no source
unknown
154.8µs
1.17×
Reported
license unknown · no source
unknown
166.3µs
1.25×
Reported
license unknown · no source
unknown
169.5µs
1.28×
Reported
license unknown · no source
unknown
195.5µs
1.47×
Reported
license unknown · no source
unknown
253.9µs
1.91×
Reported
license unknown · no source
Show all 34 implementations ›
unknown
687.7µs
5.18×
Reported
license unknown · no source
unknown
825.8µs
6.23×
Reported
license unknown · no source
unknown
3.15ms
23.76×
Reported
license unknown · no source
unknown
149.7µs
1.13×
Reported
license unknown · no source
unknown
149.3µs
1.13×
Reported
license unknown · no source
unknown
155.8µs
1.17×
Reported
license unknown · no source
unknown
171.1µs
1.29×
Reported
license unknown · no source
unknown
172.1µs
1.30×
Reported
license unknown · no source
unknown
177.1µs
1.34×
Reported
license unknown · no source
unknown
180.4µs
1.36×
Reported
license unknown · no source
unknown
181.6µs
1.37×
Reported
license unknown · no source
unknown
183.1µs
1.38×
Reported
license unknown · no source
unknown
176.8µs
1.33×
Reported
license unknown · no source
unknown
191.0µs
1.44×
Reported
license unknown · no source
unknown
196.9µs
1.48×
Reported
license unknown · no source
unknown
204.3µs
1.54×
Reported
license unknown · no source
unknown
257.9µs
1.94×
Reported
license unknown · no source
unknown
268.3µs
2.02×
Reported
license unknown · no source
unknown
732.1µs
5.52×
Reported
license unknown · no source
unknown
897.9µs
6.77×
Reported
license unknown · no source
unknown
940.7µs
7.09×
Reported
license unknown · no source
unknown
1.46ms
10.98×
Reported
license unknown · no source
unknown
1.68ms
12.69×
Reported
license unknown · no source
unknown
140.7µs
1.06×
Reported
license unknown · no source
unknown
142.1µs
1.07×
Reported
license unknown · no source
unknown
142.6µs
1.08×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesfp32 [batch_size, num_frames, num_patches_per_frame, hidden_size]
qkv_weightfp32 [qkv_hidden, hidden_size]
qkv_biasfp32 [qkv_hidden]
out_weightfp32 [hidden_size, hidden_size]
out_biasfp32 [hidden_size]
temporal_freqsfp32 [half_head_dim]
spatial_freqsfp32 [half_head_dim]
scalefp32 scalar
outputfp32 [batch_size, num_frames, num_patches_per_frame, hidden_size]
Axes and behavior
head_dimconstant = 64
batch_sizevariable
num_framesvariable
qkv_hiddenqkv_hidden = hidden_size * 3
hidden_sizeconstant = 1024
half_head_dimhalf_head_dim = head_dim // 2
num_attention_headsconstant = 16
num_patches_per_framevariable
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias027_video_spatial_attention_with_rope_3dmodelwan2-2-t2v-a14bsha256eca7aed0687a…
Sources: NVIDIA SOL-ExecBench (2026-08-23)last observed 2026-08-23How records are decidedJSON