Skip to content
KernelIndex
Search⌘K
Sign in

SAM hq vision encoder window partition attention

22 eligible runs
vision

SAM-HQ's windowed attention mechanism that fuses window partitioning, multi-head attention with 2D relative position embeddings, window unpartitioning, residual connections, layer normalization, and MLP block. Processes image patches in 14x14 windows for efficient local attention computation.

Fastest reported
2.24msmean · 88% of SOL
RIAC teamLicense unknown · unknown

Reported evidence · last observed 2026-07-30. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencymlp_dim = 3072 · head_dim = 64 · hidden_size = 768 · window_size = 14 · qkv_out_size = 2304 · rel_pos_size = 27 · num_attention_heads = 1216 cases
height
width
batch_size
dtype
64644fp3264648fp32646416fp321281284fp3212812816fp321311311fp321923842fp322242241fp322562562fp322562568fp322932931fp323363361fp324494491fp325125121fp326136131fp32102410241fp32
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 12.24 msRIAC team17NVIDIA B200 · env 22.53 msdoubleAI5

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 · 17 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
RIAC team
2.24ms88% SOL
1.00×
Reported · license unknown · no source
2026-07-30

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

no sourcelicense unknownno install recipeRun detail →
2
doubleAI
2.67ms84% SOL
1.19×
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 →
3
Amir M. Mir | SF Tensor
2.67ms84% SOL
1.19×
Reported · license unknown · no source
2026-07-18

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

no sourcelicense unknownno install recipeRun detail →
4
ac4k
2.72ms84% SOL
1.21×
Reported · license unknown · no source
2026-08-03

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

no sourcelicense unknownno install recipeRun detail →
5
Geometric
3.02ms82% SOL
1.35×
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 →
6
Geometric
3.28ms80% SOL
1.47×
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 →
7
Recursive
3.34ms80% SOL
1.49×
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
Proud Wren @ NVIDIA
3.34ms80% SOL
1.49×
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
3.44ms79% SOL
1.54×
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 →
10
Geometric
3.44ms79% SOL
1.54×
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 →
11
Hyra
3.93ms76% SOL
1.75×
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 →
12
Geometric
4.68ms71% SOL
2.09×
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 →
13
Geometric
4.68ms71% SOL
2.09×
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 →
14
Geometric
6.13ms64% SOL
2.74×
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
Cursor
6.54ms63% SOL
2.92×
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
Amir M. Mir | SF Tensor
12.1ms46% SOL
5.39×
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 →
17
Geometric
41.1ms19% SOL
18.35×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
RIAC team
unknown
2.24ms
1.00×
Reported
license unknown · no source
doubleAI
unknown
2.53ms
1.13×
Reported
license unknown · no source
doubleAI
unknown
2.67ms
1.19×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
2.67ms
1.19×
Reported
license unknown · no source
ac4k
unknown
2.72ms
1.21×
Reported
license unknown · no source
Geometric
unknown
3.02ms
1.35×
Reported
license unknown · no source
Recursive
unknown
3.26ms
1.46×
Reported
license unknown · no source
Proud Wren @ NVIDIA
unknown
3.34ms
1.49×
Reported
license unknown · no source
Show all 22 implementations ›
Hyra
unknown
3.93ms
1.75×
Reported
license unknown · no source
Cursor
unknown
6.54ms
2.92×
Reported
license unknown · no source
doubleAI
unknown
3.04ms
1.36×
Reported
license unknown · no source
Geometric
unknown
3.28ms
1.47×
Reported
license unknown · no source
Recursive
unknown
3.34ms
1.49×
Reported
license unknown · no source
Geometric
unknown
3.44ms
1.54×
Reported
license unknown · no source
Geometric
unknown
3.44ms
1.54×
Reported
license unknown · no source
Geometric
unknown
4.68ms
2.09×
Reported
license unknown · no source
Geometric
unknown
4.68ms
2.09×
Reported
license unknown · no source
Geometric
unknown
6.13ms
2.74×
Reported
license unknown · no source
Cursor
unknown
6.78ms
3.03×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
11.8ms
5.26×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
12.1ms
5.39×
Reported
license unknown · no source
Geometric
unknown
41.1ms
18.35×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesfp32 [batch_size, height, width, hidden_size]
qkv_weightfp32 [qkv_out_size, hidden_size]
qkv_biasfp32 [qkv_out_size]
proj_weightfp32 [hidden_size, hidden_size]
proj_biasfp32 [hidden_size]
rel_pos_hfp32 [rel_pos_size, head_dim]
rel_pos_wfp32 [rel_pos_size, head_dim]
layer_norm1_weightfp32 [hidden_size]
layer_norm1_biasfp32 [hidden_size]
layer_norm2_weightfp32 [hidden_size]
layer_norm2_biasfp32 [hidden_size]
mlp_lin1_weightfp32 [mlp_dim, hidden_size]
mlp_lin1_biasfp32 [mlp_dim]
mlp_lin2_weightfp32 [hidden_size, mlp_dim]
mlp_lin2_biasfp32 [hidden_size]
layer_norm_epsfp32 scalar
outputfp32 [batch_size, height, width, hidden_size]
Axes and behavior
widthvariable
heightvariable
mlp_dimconstant = 3072
head_dimconstant = 64
batch_sizevariable
hidden_sizeconstant = 768
window_sizeconstant = 14
qkv_out_sizeconstant = 2304
rel_pos_sizeconstant = 27
num_attention_headsconstant = 12
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias077_sam_hq_vision_encoder_window_partition_attentionmodelsam-hq-vit-basesha256fedaec3e5ce8…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON