Conjectures.io
Tensor kernels

Entry #114

Completed
Hotkey
5D4arF…FdtT3L
Sent
12 Nov 2026, 15:51 UTC
Source
cb3fac24f297

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 #50Done2 of 2 runs1.004×Generated 1.023×, real models 0.985×Above the barGain lower bound -7.6%Guard passed
MLPvs #44Done2 of 2 runs1.053×Generated 1.008×, real models 1.100×Above the barGain lower bound -32.2%Guard passed
ResNetvs #59Done2 of 2 runs0.987×Generated 0.976×, real models 0.999×Above the barGain lower bound -62.9%Guard passed
MobileNetvs #53Done2 of 2 runs1.125×Generated 1.216×, real models 1.041×Above the barGain lower bound -26.1%Guard passed
Normalized MLPvs #48Done2 of 2 runs1.108×Generated 0.931×, real models 1.318×Above the barGain lower bound -32.1%Guard passed
Mixervs #58Done2 of 2 runs0.988×Generated 0.997×, real models 0.979×Above the barGain lower bound -24.5%Guard passed
Recurrentvs #57Done2 of 2 runs0.944×Generated 0.952×, real models 0.936×QualifiesGain lower bound 3.7%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 1.023× · real models 0.985×

MachineTime
  1. prod-amd64Seed 1501147980.981×Done1 attempt0.009× of the baseline kernels' time
  2. prod-arm64Seed 1501147981.027×Done1 attempt0.011× of the baseline kernels' time

By model and inputGuard at 1.100×

MLP

  • Batch 1, normal1.231×
  • Batch 1, sparse1.243×
  • Batch 1, wide1.296×

Mixer

  • Batch 1, normal1.024×
  • Batch 1, sparse1.204×
  • Batch 1, wide1.041×

ResNet

  • Batch 1, normal1.007×
  • Batch 1, sparse1.012×
  • Batch 1, wide0.940×

MobileNet

  • Batch 1, normal1.022×
  • Batch 1, sparse1.045×
  • Batch 1, wide0.992×

Recurrent

  • Batch 1, normal0.850×
  • Batch 1, sparse0.825×
  • Batch 1, wide0.795×

Normalized MLP

  • Batch 1, normal1.197×
  • Batch 1, sparse1.222×
  • Batch 1, wide1.156×

Real models

  • appliance-energy-mixer on Appliance Energy Prediction0.778×
  • compact-mixer on digits0.847×
  • compact-mixer on fashion0.960×
  • compact-mixer on mnist1.041×
  • compact-mlp on digits0.974×
  • compact-mlp on fashion2.344×
  • compact-mlp on mnist1.104×
  • compact-mobilenet on digits0.776×
  • compact-mobilenet on fashion0.911×
  • compact-mobilenet on mnist0.894×
  • compact-normalized_mlp on digits1.019×
  • compact-normalized_mlp on fashion1.213×
  • compact-normalized_mlp on mnist1.365×
  • compact-resnet on digits0.835×
  • compact-resnet on fashion0.922×
  • compact-resnet on mnist0.933×
  • appliance-gru on appliance-energy0.848×
  • mitbih-gru on mitbih0.765×
  • mnist-rows-lstm on mnist-rows0.556×
  • speech-commands-gru on speech-commands0.818×
  • speech-commands-lstm-gru on speech-commands0.829×
  • spoken-arabic-digit-lstm on spoken-arabic-digit0.886×
  • tiny-shakespeare-lstm on tiny-shakespeare0.615×
  • uci-har-gru on uci-har0.788×
  • uci-har-lstm on uci-har0.859×
  • uci-har-mixer on UCI-HAR0.997×
  • uci-har-mlp on UCI-HAR1.005×
  • uci-har-mobilenet on UCI-HAR1.195×
  • uci-har-normalized_mlp on UCI-HAR1.012×
  • uci-har-resnet on UCI-HAR1.225×
MLP2 runs

RunsGenerated 1.008× · real models 1.100×

MachineTime
  1. prod-amd64Seed 1501147981.104×Done1 attempt0.014× of the baseline kernels' time
  2. prod-arm64Seed 1501147981.004×Done1 attempt0.018× of the baseline kernels' time

By model and inputGuard at 1.100×

MLP

  • Batch 1, normal0.987×
  • Batch 1, sparse1.019×
  • Batch 1, wide1.001×

Real models

  • compact-mlp on digits0.964×
  • compact-mlp on fashion1.138×
  • compact-mlp on mnist1.339×
  • uci-har-mlp on UCI-HAR0.996×
ResNet2 runs

RunsGenerated 0.976× · real models 0.999×

MachineTime
  1. prod-amd64Seed 1501147980.975×Done1 attempt0.008× of the baseline kernels' time
  2. prod-arm64Seed 1501147980.999×Done1 attempt0.008× of the baseline kernels' time

By model and inputGuard at 1.100×

ResNet

  • Batch 1, normal0.956×
  • Batch 1, sparse1.004×
  • Batch 1, wide0.998×

Real models

  • compact-resnet on digits1.005×
  • compact-resnet on fashion1.002×
  • compact-resnet on mnist0.977×
  • uci-har-resnet on UCI-HAR1.011×
MobileNet2 runs

RunsGenerated 1.216× · real models 1.041×

MachineTime
  1. prod-amd64Seed 1501147981.023×Done1 attempt0.012× of the baseline kernels' time
  2. prod-arm64Seed 1501147981.238×Done1 attempt0.013× of the baseline kernels' time

By model and inputGuard at 1.100×

MobileNet

  • Batch 1, normal1.443×
  • Batch 1, sparse1.058×
  • Batch 1, wide0.990×

Real models

  • compact-mobilenet on digits0.897×
  • compact-mobilenet on fashion1.020×
  • compact-mobilenet on mnist1.070×
  • uci-har-mobilenet on UCI-HAR1.199×
Normalized MLP2 runs

RunsGenerated 0.931× · real models 1.318×

MachineTime
  1. prod-amd64Seed 1501147980.961×Done1 attempt0.008× of the baseline kernels' time
  2. prod-arm64Seed 1501147981.277×Done1 attempt0.010× of the baseline kernels' time

By model and inputGuard at 1.100×

Normalized MLP

  • Batch 1, normal0.850×
  • Batch 1, sparse0.971×
  • Batch 1, wide0.995×

Real models

  • compact-normalized_mlp on digits1.082×
  • compact-normalized_mlp on fashion1.224×
  • compact-normalized_mlp on mnist1.764×
  • uci-har-normalized_mlp on UCI-HAR1.290×
Mixer2 runs

RunsGenerated 0.997× · real models 0.979×

MachineTime
  1. prod-amd64Seed 1501147980.981×Done1 attempt0.004× of the baseline kernels' time
  2. prod-arm64Seed 1501147980.996×Done1 attempt0.005× of the baseline kernels' time

By model and inputGuard at 1.100×

Mixer

  • Batch 1, normal1.009×
  • Batch 1, sparse1.113×
  • Batch 1, wide0.943×

Real models

  • appliance-energy-mixer on Appliance Energy Prediction0.865×
  • compact-mixer on digits0.996×
  • compact-mixer on fashion1.030×
  • compact-mixer on mnist1.001×
  • uci-har-mixer on UCI-HAR1.011×
Recurrent2 runs

RunsGenerated 0.952× · real models 0.936×

MachineTime
  1. prod-amd64Seed 1501147980.952×Done1 attempt0.009× of the baseline kernels' time
  2. prod-arm64Seed 1501147980.935×Done1 attempt0.011× of the baseline kernels' time

By model and inputGuard at 1.100×

Recurrent

  • Batch 1, normal0.975×
  • Batch 1, sparse0.964×
  • Batch 1, wide0.957×

Real models

  • appliance-gru on appliance-energy1.020×
  • mitbih-gru on mitbih0.957×
  • mnist-rows-lstm on mnist-rows0.792×
  • speech-commands-gru on speech-commands0.976×
  • speech-commands-lstm-gru on speech-commands0.929×
  • spoken-arabic-digit-lstm on spoken-arabic-digit0.969×
  • tiny-shakespeare-lstm on tiny-shakespeare0.901×
  • uci-har-gru on uci-har0.951×
  • uci-har-lstm on uci-har0.944×