Skip to content
KernelIndex
Search⌘K
Sign in

Video latent VAE encoder downsampling

24 eligible runs
video

VAE encoder downsampling block for video-to-latent compression. Converts raw video frames to compressed latent representations through spatial and temporal downsampling with 3D convolutions, group normalization, SiLU activations, and residual connections.

Fastest reported
3.99msmean · 81% of SOL
Amir M. Mir | SF TensorLicense unknown · unknown

Reported evidence · last observed 2026-08-23. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencynum_groups = 32 · groups_div2 = 16 · groups_div4 = 8 · hidden_div2 = 512 · hidden_div4 = 256 · hidden_size = 1024 · in_channels = 3 · latent_channels = 1616 cases
batch_size
height
width
frames
dtype
112812848fp32119219216fp32125625616fp32125625624fp32125625632fp3213843848fp32138438416fp32144844816fp32212812824fp32219219216fp3222562568fp32412812816fp3242562568fp32812812816fp323264648fp326464648fp32
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 13.99 msAmir M. Mir | SF Tensor16NVIDIA B200 · env 24.30 msRecursive8

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 · 16 results · last observed 2026-08-23Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
Amir M. Mir | SF Tensor
3.99ms81% SOL
1.00×
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 →
2
RIAC as well
4.13ms80% SOL
1.04×
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 →
3
Proud Wren @ NVIDIA
4.15ms80% SOL
1.04×
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
ac4k
4.33ms78% SOL
1.09×
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 →
5
Hyra
4.39ms78% SOL
1.10×
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 →
6
Recursive
4.53ms77% 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 →
7
Proud Wren @ NVIDIA
4.54ms77% 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 →
8
Amir M. Mir | SF Tensor
4.70ms76% SOL
1.18×
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 →
9
Geometric
5.58ms70% SOL
1.40×
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 →
10
doubleAI
5.67ms69% SOL
1.42×
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
Geometric
6.14ms67% SOL
1.54×
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 →
12
Amir M. Mir | SF Tensor
7.67ms59% SOL
1.92×
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 →
13
Cursor
11.2ms47% SOL
2.81×
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
gnehiz
16.4ms36% SOL
4.11×
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 →
15
Geometric
21.5ms29% SOL
5.39×
Reported · license unknown · no source
2026-07-25

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

no sourcelicense unknownno install recipeRun detail →
16
Amir M. Mir | SF Tensor
23.6ms31% SOL
5.92×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
Amir M. Mir | SF Tensor
unknown
3.99ms
1.00×
Reported
license unknown · no source
RIAC as well
unknown
4.13ms
1.04×
Reported
license unknown · no source
Hyra
unknown
4.39ms
1.10×
Reported
license unknown · no source
doubleAI
unknown
4.82ms
1.21×
Reported
license unknown · no source
Geometric
unknown
5.58ms
1.40×
Reported
license unknown · no source
Cursor
unknown
10.3ms
2.59×
Reported
license unknown · no source
gnehiz
unknown
15.9ms
3.99×
Reported
license unknown · no source
Proud Wren @ NVIDIA
unknown
4.42ms
1.11×
Reported
license unknown · no source
Show all 24 implementations ›
Recursive
unknown
4.53ms
1.14×
Reported
license unknown · no source
Proud Wren @ NVIDIA
unknown
4.54ms
1.14×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
4.70ms
1.18×
Reported
license unknown · no source
doubleAI
unknown
5.05ms
1.27×
Reported
license unknown · no source
doubleAI
unknown
5.67ms
1.42×
Reported
license unknown · no source
Geometric
unknown
6.14ms
1.54×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
7.67ms
1.92×
Reported
license unknown · no source
Cursor
unknown
11.2ms
2.81×
Reported
license unknown · no source
gnehiz
unknown
16.4ms
4.11×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
23.2ms
5.80×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
23.6ms
5.92×
Reported
license unknown · no source
Geometric
unknown
21.5ms
5.39×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
41.7ms
10.45×
Reported
license unknown · no source
Proud Wren @ NVIDIA
unknown
4.15ms
1.04×
Reported
license unknown · no source
Recursive
unknown
4.30ms
1.08×
Reported
license unknown · no source
ac4k
unknown
4.33ms
1.09×
Reported
license unknown · no source

Semantics

Inputs and outputs
xfp32 [batch_size, in_channels, frames, height, width]
input_conv_weightfp32 [hidden_div4, in_channels, 3, 3, 3]
input_conv_biasfp32 [hidden_div4]
down1_gn_weightfp32 [hidden_div4]
down1_gn_biasfp32 [hidden_div4]
down1_conv_weightfp32 [hidden_div2, hidden_div4, 3, 4, 4]
down1_conv_biasfp32 [hidden_div2]
down1_res_weightfp32 [hidden_div2, hidden_div4, 1, 1, 1]
down1_res_biasfp32 [hidden_div2]
down2_gn_weightfp32 [hidden_div2]
down2_gn_biasfp32 [hidden_div2]
down2_conv_weightfp32 [hidden_size, hidden_div2, 3, 4, 4]
down2_conv_biasfp32 [hidden_size]
down2_res_weightfp32 [hidden_size, hidden_div2, 1, 1, 1]
down2_res_biasfp32 [hidden_size]
down3_gn_weightfp32 [hidden_size]
down3_gn_biasfp32 [hidden_size]
down3_conv_weightfp32 [hidden_size, hidden_size, 3, 4, 4]
down3_conv_biasfp32 [hidden_size]
down3_res_weightfp32 [hidden_size, hidden_size, 1, 1, 1]
down3_res_biasfp32 [hidden_size]
latent_gn_weightfp32 [hidden_size]
latent_gn_biasfp32 [hidden_size]
latent_conv_weightfp32 [latent_channels, hidden_size, 3, 3, 3]
latent_conv_biasfp32 [latent_channels]
epsfp32 scalar
outputfp32 [batch_size, latent_channels, frames_div4, height_div8, width_div8]
Axes and behavior
widthvariable
framesvariable
heightvariable
batch_sizevariable
num_groupsconstant = 32
width_div2width_div2 = width // 2
width_div4width_div4 = width // 4
width_div8width_div8 = width // 8
frames_div2frames_div2 = frames // 2
frames_div4frames_div4 = frames // 4
groups_div2groups_div2 = num_groups // 2
groups_div4groups_div4 = num_groups // 4
height_div2height_div2 = height // 2
height_div4height_div4 = height // 4
height_div8height_div8 = height // 8
hidden_div2hidden_div2 = hidden_size // 2
hidden_div4hidden_div4 = hidden_size // 4
hidden_sizeconstant = 1024
in_channelsconstant = 3
latent_channelsconstant = 16
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias023_video_latent_vae_encoder_downsamplingmodelwan2-2-t2v-a14bsha256bcdce832dec4…
Sources: NVIDIA SOL-ExecBench (2026-08-23)last observed 2026-08-23How records are decidedJSON