/dlprof

Deep Learning Profiling Project - EE6040 @ Columbia University

Primary LanguageJupyter Notebook

GPU Profiling For Tensorflow Performance

As our final project, we decided to create a compilation of Tensorflow best practices using GPU profiling tools. The final project, presented as a series of blog posts, can be found https://aistein.github.io/dlprof/.

Director Descriptions

  1. DeepCoNN - This directory contains examples involving the Joint Modeling of User and Item Review model.
  2. docs - This directory contains the source for the blog
  3. images - This directory contains images presented as a project update on 4/15/18
  4. intel_mkl - This directory contains all of the necessary code and scripts to build tensorflow from scratch with various different settings
  5. misc - This directory contains miscellaneous data files used to generate graphs or present results
  6. mnist_data_format - This directory contains examples testing data formats for convolutional networks
  7. mnist_xla - This directory contains tests and configurations to compile tensorflow from scratch using XLA and test it
  8. preprocessing - This directory contains code necessary for preprocessing text data for the DeepCoNN models
  9. report - This directory contains the latex source for our final report submission

Michael Alvarino - maa2282@columbia.edu Alexander Stein - as5281@columbia.edu