Skip to content

Instantly share code, notes, and snippets.

@ypwhs

ypwhs/FLOPS_Test.ipynb

Last active November 24, 2018 18:12
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save ypwhs/a156353cb19e780e323be054821af015 to your computer and use it in GitHub Desktop.
Save ypwhs/a156353cb19e780e323be054821af015 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
README.md

PyTorch Benchmark

Environment

  • torch version: 1.0.0a0+ff608a9
  • CUDA version: 10.0.130
  • DISTRIB_DESCRIPTION="Ubuntu 16.04 LTS"

Layer config

Linear

  • Input shape: torch.Size([1, 4096, 4096])
  • Op: Linear(in_features=4096, out_features=4096, bias=False)
  • 128.000GFLOP

Conv2d

  • Input shape: torch.Size([1, 256, 256, 256])
  • Op: Conv2d(256, 256, kernel_size=(8, 8), stride=(1, 1), bias=False)
  • 484.383GFLOP

Result

GPU Model FP32 Linear FP16 Linear Linear Speedup FP32 Conv2d FP16 Conv2d Conv2d Speedup
TITAN V 12.9TFLOPS 77.4TFLOPS 598.61% 8.71TFLOPS 76.4TFLOPS 875.61%
RTX 2080 Ti 12.7TFLOPS 53.4TFLOPS 419.31% 6.20TFLOPS 51.9TFLOPS 750.22%
RTX 2080 9.06TFLOPS 36.2TFLOPS 398.94% 4.95TFLOPS 35.8TFLOPS 721.24%
GTX 1080 Ti 10.5TFLOPS 10.3TFLOPS 98.74% 6.29TFLOPS 7.14TFLOPS 108.55%
Display the source blob
Display the rendered blob
Raw
FLOPS_Test.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment