Conjectures.io
Tensor kernels

Entry #26

Completed
Hotkey
5FxfWH…FfnvBb
Sent
11 Nov 2026, 02:25 UTC
Source
df7425de5eb9

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 #29Done2 of 2 runs0.942×Generated 0.998×, real models 0.888×Not significantGain lower bound -1.3%Guard passed
MLPvs #1Done2 of 2 runs1.077×Generated 0.925×, real models 1.253×Above the barGain lower bound -33.4%Guard passed
ResNetvs #32Done2 of 2 runs1.011×Generated 1.028×, real models 0.995×Above the barGain lower bound -3.4%Guard passed
MobileNetvs #33Done2 of 2 runs1.000×Generated 1.035×, real models 0.965×Above the barGain lower bound -7.7%Guard passed
Normalized MLPvs #6Done2 of 2 runs0.989×Generated 0.997×, real models 0.981×Above the barGain lower bound -23.3%Guard passed
Mixervs #24Done2 of 2 runs1.015×Generated 1.021×, real models 1.008×Above the barGain lower bound -15.5%Guard passed
Recurrentvs #25Done2 of 2 runs0.896×Generated 1.009×, real models 0.796×QualifiesGain lower bound 2.6%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 0.998× · real models 0.888×

MachineTime
  1. prod-amd64Seed 3809890240.953×Done1 attempt0.012× of the baseline kernels' time
  2. prod-arm64Seed 3809890240.931×Done1 attempt0.014× of the baseline kernels' time

By model and inputGuard at 1.100×

MLP

  • Batch 1, normal0.932×
  • Batch 1, sparse0.996×
  • Batch 1, wide0.961×

Mixer

  • Batch 1, normal1.078×
  • Batch 1, sparse1.065×
  • Batch 1, wide1.085×

ResNet

  • Batch 1, normal1.008×
  • Batch 1, sparse1.036×
  • Batch 1, wide0.986×

MobileNet

  • Batch 1, normal0.981×
  • Batch 1, sparse1.045×
  • Batch 1, wide0.988×

Recurrent

  • Batch 1, normal1.033×
  • Batch 1, sparse1.029×
  • Batch 1, wide1.028×

Normalized MLP

  • Batch 1, normal1.094×
  • Batch 1, sparse1.077×
  • Batch 1, wide1.013×

Real models

  • appliance-energy-mixer on Appliance Energy Prediction1.653×
  • compact-mixer on digits1.575×
  • compact-mixer on fashion1.040×
  • compact-mixer on mnist0.895×
  • compact-mlp on digits1.278×
  • compact-mlp on fashion1.266×
  • compact-mlp on mnist0.982×
  • compact-mobilenet on digits0.887×
  • compact-mobilenet on fashion0.741×
  • compact-mobilenet on mnist0.687×
  • compact-normalized_mlp on digits0.994×
  • compact-normalized_mlp on fashion1.068×
  • compact-normalized_mlp on mnist1.012×
  • compact-resnet on digits0.954×
  • compact-resnet on fashion0.847×
  • compact-resnet on mnist0.783×
  • appliance-gru on appliance-energy1.115×
  • mitbih-gru on mitbih0.994×
  • mnist-rows-lstm on mnist-rows0.682×
  • speech-commands-gru on speech-commands1.276×
  • speech-commands-lstm-gru on speech-commands1.034×
  • spoken-arabic-digit-lstm on spoken-arabic-digit0.611×
  • tiny-shakespeare-lstm on tiny-shakespeare0.713×
  • uci-har-gru on uci-har1.094×
  • uci-har-lstm on uci-har0.289×
  • uci-har-mixer on UCI-HAR1.042×
  • uci-har-mlp on UCI-HAR1.056×
  • uci-har-mobilenet on UCI-HAR0.269×
  • uci-har-normalized_mlp on UCI-HAR0.997×
  • uci-har-resnet on UCI-HAR0.484×
MLP2 runs

RunsGenerated 0.925× · real models 1.253×

MachineTime
  1. prod-amd64Seed 3809890240.953×Done1 attempt0.013× of the baseline kernels' time
  2. prod-arm64Seed 3809890241.217×Done1 attempt0.016× of the baseline kernels' time

By model and inputGuard at 1.100×

MLP

  • Batch 1, normal0.917×
  • Batch 1, sparse0.865×
  • Batch 1, wide0.969×

Real models

  • compact-mlp on digits1.968×
  • compact-mlp on fashion0.927×
  • compact-mlp on mnist1.497×
  • uci-har-mlp on UCI-HAR0.904×
ResNet2 runs

RunsGenerated 1.028× · real models 0.995×

MachineTime
  1. prod-amd64Seed 3809890241.025×Done1 attempt0.009× of the baseline kernels' time
  2. prod-arm64Seed 3809890240.999×Done1 attempt0.010× of the baseline kernels' time

By model and inputGuard at 1.100×

ResNet

  • Batch 1, normal1.030×
  • Batch 1, sparse1.010×
  • Batch 1, wide1.028×

Real models

  • compact-resnet on digits1.039×
  • compact-resnet on fashion0.980×
  • compact-resnet on mnist0.962×
  • uci-har-resnet on UCI-HAR1.000×
MobileNet2 runs

RunsGenerated 1.035× · real models 0.965×

MachineTime
  1. prod-amd64Seed 3809890241.027×Done1 attempt0.014× of the baseline kernels' time
  2. prod-arm64Seed 3809890240.973×Done1 attempt0.014× of the baseline kernels' time

By model and inputGuard at 1.100×

MobileNet

  • Batch 1, normal1.025×
  • Batch 1, sparse1.063×
  • Batch 1, wide1.006×

Real models

  • compact-mobilenet on digits0.998×
  • compact-mobilenet on fashion0.954×
  • compact-mobilenet on mnist0.939×
  • uci-har-mobilenet on UCI-HAR0.971×
Normalized MLP2 runs

RunsGenerated 0.997× · real models 0.981×

MachineTime
  1. prod-amd64Seed 3809890241.073×Done1 attempt0.010× of the baseline kernels' time
  2. prod-arm64Seed 3809890240.912×Done1 attempt0.010× of the baseline kernels' time

By model and inputGuard at 1.100×

Normalized MLP

  • Batch 1, normal1.006×
  • Batch 1, sparse1.044×
  • Batch 1, wide1.000×

Real models

  • compact-normalized_mlp on digits1.048×
  • compact-normalized_mlp on fashion0.931×
  • compact-normalized_mlp on mnist1.004×
  • uci-har-normalized_mlp on UCI-HAR0.947×
Mixer2 runs

RunsGenerated 1.021× · real models 1.008×

MachineTime
  1. prod-amd64Seed 3809890241.034×Done1 attempt0.005× of the baseline kernels' time
  2. prod-arm64Seed 3809890240.995×Done1 attempt0.005× of the baseline kernels' time

By model and inputGuard at 1.100×

Mixer

  • Batch 1, normal0.982×
  • Batch 1, sparse1.040×
  • Batch 1, wide0.996×

Real models

  • appliance-energy-mixer on Appliance Energy Prediction1.034×
  • compact-mixer on digits1.082×
  • compact-mixer on fashion0.953×
  • compact-mixer on mnist0.960×
  • uci-har-mixer on UCI-HAR1.016×
Recurrent2 runs

RunsGenerated 1.009× · real models 0.796×

MachineTime
  1. prod-amd64Seed 3809890240.893×Done1 attempt0.042× of the baseline kernels' time
  2. prod-arm64Seed 3809890240.899×Done1 attempt0.044× of the baseline kernels' time

By model and inputGuard at 1.100×

Recurrent

  • Batch 1, normal1.024×
  • Batch 1, sparse1.027×
  • Batch 1, wide1.022×

Real models

  • appliance-gru on appliance-energy1.109×
  • mitbih-gru on mitbih0.972×
  • mnist-rows-lstm on mnist-rows0.680×
  • speech-commands-gru on speech-commands1.271×
  • speech-commands-lstm-gru on speech-commands1.037×
  • spoken-arabic-digit-lstm on spoken-arabic-digit0.606×
  • tiny-shakespeare-lstm on tiny-shakespeare0.712×
  • uci-har-gru on uci-har1.081×
  • uci-har-lstm on uci-har0.284×