Conjectures.io
Tensor kernels

Entry #7

Completed
Hotkey
5DcV46…wJGDCh
Sent
10 Nov 2026, 19:57 UTC
Source
9239f791cd78

Progress

  1. Queued
  2. Proof
  3. Planned
  4. Timing
  5. Done

Done means every board has a decision. It does not mean the entry took a board over.

On each board

BoardRunsTimeDecision
Overallvs #11Done2 of 2 runs1.466×Generated 1.914×, real models 1.123×Above the barGain lower bound -60.7%Guard passed
MLPvs #1Done2 of 2 runs0.975×Generated 1.006×, real models 0.944×Above the barGain lower bound -9.5%Guard passed
ResNetvs #12Done2 of 2 runs0.490×Generated 0.496×, real models 0.483×QualifiesGain lower bound 48.6%Guard passed
MobileNetvs #13Done2 of 2 runs0.480×Generated 0.463×, real models 0.498×QualifiesGain lower bound 50.9%Guard passed
Normalized MLPvs #6Done2 of 2 runs41.799×Generated 48.847×, real models 35.768×Above the barGain lower bound -5375.1%Guard passed
Mixervs #7Done2 of 2 runs117.699×Generated 165.183×, real models 83.864×Above the barGain lower bound -14514.8%Guard passed
Recurrentvs #14Done2 of 2 runs0.999×Generated 0.999×, real models 0.999×Above the barGain lower bound -0.3%Guard passed

Download Kernel.lean

Every run

Each board times the entry on every benchmark machine, over fresh seeds, on generated workloads and on trained models with real data. A run's time combines the two suites by their published weights; the bars below split it by model, batch and input, and by real model.

Overall2 runs

RunsGenerated 1.914× · real models 1.123×

MachineTime
  1. prod-amd64Seed 6754662961.414×Done1 attempt0.207× of the baseline kernels' time
  2. prod-arm64Seed 6754662961.521×Done1 attempt0.219× of the baseline kernels' time

By model and inputGuard at 1.100×

MLP

  • Batch 1, normal0.955×
  • Batch 1, sparse0.963×
  • Batch 1, wide0.890×

Mixer

  • Batch 1, normal137.870×
  • Batch 1, sparse134.926×
  • Batch 1, wide168.161×

ResNet

  • Batch 1, normal0.500×
  • Batch 1, sparse0.494×
  • Batch 1, wide0.492×

MobileNet

  • Batch 1, normal0.384×
  • Batch 1, sparse0.388×
  • Batch 1, wide0.384×

Recurrent

  • Batch 1, normal0.044×
  • Batch 1, sparse0.045×
  • Batch 1, wide0.045×

Normalized MLP

  • Batch 1, normal46.900×
  • Batch 1, sparse42.713×
  • Batch 1, wide44.664×

Real models

  • appliance-energy-mixer on Appliance Energy Prediction171.338×
  • compact-mixer on digits64.669×
  • compact-mixer on fashion45.271×
  • compact-mixer on mnist32.837×
  • compact-mlp on digits0.771×
  • compact-mlp on fashion0.908×
  • compact-mlp on mnist0.797×
  • compact-mobilenet on digits0.227×
  • compact-mobilenet on fashion0.203×
  • compact-mobilenet on mnist0.119×
  • compact-normalized_mlp on digits51.499×
  • compact-normalized_mlp on fashion17.826×
  • compact-normalized_mlp on mnist20.070×
  • compact-resnet on digits0.251×
  • compact-resnet on fashion0.097×
  • compact-resnet on mnist0.070×
  • appliance-gru on appliance-energy0.046×
  • mitbih-gru on mitbih0.040×
  • mnist-rows-lstm on mnist-rows0.032×
  • speech-commands-gru on speech-commands0.054×
  • speech-commands-lstm-gru on speech-commands0.038×
  • spoken-arabic-digit-lstm on spoken-arabic-digit0.032×
  • tiny-shakespeare-lstm on tiny-shakespeare0.030×
  • uci-har-gru on uci-har0.044×
  • uci-har-lstm on uci-har0.041×
  • uci-har-mixer on UCI-HAR126.891×
  • uci-har-mlp on UCI-HAR1.085×
  • uci-har-mobilenet on UCI-HAR0.387×
  • uci-har-normalized_mlp on UCI-HAR17.465×
  • uci-har-resnet on UCI-HAR0.352×
MLP2 runs

RunsGenerated 1.006× · real models 0.944×

MachineTime
  1. prod-amd64Seed 6754662960.953×Done1 attempt0.019× of the baseline kernels' time
  2. prod-arm64Seed 6754662960.997×Done1 attempt0.018× of the baseline kernels' time

By model and inputGuard at 1.100×

MLP

  • Batch 1, normal0.990×
  • Batch 1, sparse0.969×
  • Batch 1, wide1.069×

Real models

  • compact-mlp on digits0.715×
  • compact-mlp on fashion1.058×
  • compact-mlp on mnist1.021×
  • uci-har-mlp on UCI-HAR1.031×
ResNet2 runs

RunsGenerated 0.496× · real models 0.483×

MachineTime
  1. prod-amd64Seed 6754662960.492×Done1 attempt0.141× of the baseline kernels' time
  2. prod-arm64Seed 6754662960.487×Done1 attempt0.144× of the baseline kernels' time

By model and inputGuard at 1.100×

ResNet

  • Batch 1, normal0.502×
  • Batch 1, sparse0.493×
  • Batch 1, wide0.492×

Real models

  • compact-resnet on digits0.504×
  • compact-resnet on fashion0.536×
  • compact-resnet on mnist0.557×
  • uci-har-resnet on UCI-HAR0.362×
MobileNet2 runs

RunsGenerated 0.463× · real models 0.498×

MachineTime
  1. prod-amd64Seed 6754662960.476×Done1 attempt0.233× of the baseline kernels' time
  2. prod-arm64Seed 6754662960.484×Done1 attempt0.245× of the baseline kernels' time

By model and inputGuard at 1.100×

MobileNet

  • Batch 1, normal0.462×
  • Batch 1, sparse0.463×
  • Batch 1, wide0.463×

Real models

  • compact-mobilenet on digits0.489×
  • compact-mobilenet on fashion0.522×
  • compact-mobilenet on mnist0.543×
  • uci-har-mobilenet on UCI-HAR0.443×
Normalized MLP2 runs

RunsGenerated 48.847× · real models 35.768×

MachineTime
  1. prod-amd64Seed 67546629649.343×Done1 attempt0.529× of the baseline kernels' time
  2. prod-arm64Seed 67546629635.409×Done1 attempt0.535× of the baseline kernels' time

By model and inputGuard at 1.100×

Normalized MLP

  • Batch 1, normal48.234×
  • Batch 1, sparse51.781×
  • Batch 1, wide51.492×

Real models

  • compact-normalized_mlp on digits70.595×
  • compact-normalized_mlp on fashion38.684×
  • compact-normalized_mlp on mnist32.926×
  • uci-har-normalized_mlp on UCI-HAR18.203×
Mixer2 runs

RunsGenerated 165.183× · real models 83.864×

MachineTime
  1. prod-amd64Seed 675466296119.430×Done1 attempt0.784× of the baseline kernels' time
  2. prod-arm64Seed 675466296115.992×Done1 attempt0.795× of the baseline kernels' time

By model and inputGuard at 1.100×

Mixer

  • Batch 1, normal152.735×
  • Batch 1, sparse149.549×
  • Batch 1, wide198.321×

Real models

  • appliance-energy-mixer on Appliance Energy Prediction204.547×
  • compact-mixer on digits63.330×
  • compact-mixer on fashion90.451×
  • compact-mixer on mnist26.819×
  • uci-har-mixer on UCI-HAR132.013×
Recurrent2 runs

RunsGenerated 0.999× · real models 0.999×

MachineTime
  1. prod-amd64Seed 6754662961.005×Done1 attempt0.367× of the baseline kernels' time
  2. prod-arm64Seed 6754662960.993×Done1 attempt0.363× of the baseline kernels' time

By model and inputGuard at 1.100×

Recurrent

  • Batch 1, normal0.998×
  • Batch 1, sparse1.000×
  • Batch 1, wide0.998×

Real models

  • appliance-gru on appliance-energy0.996×
  • mitbih-gru on mitbih0.999×
  • mnist-rows-lstm on mnist-rows1.000×
  • speech-commands-gru on speech-commands0.996×
  • speech-commands-lstm-gru on speech-commands0.997×
  • spoken-arabic-digit-lstm on spoken-arabic-digit1.005×
  • tiny-shakespeare-lstm on tiny-shakespeare1.003×
  • uci-har-gru on uci-har0.996×
  • uci-har-lstm on uci-har1.001×