Almost two months have passed since the official release of PyTorch 1.0… at their ‘convention’ I believe, it is time to do a systematic hands on test of what it does, then try the usual standarts and, maybe, something more fun as the last part of this run. I decided to document the experiment here, because the contemporary employers require ‘blog posts and examples of code on GitHub or Kaggle’ in their ‘required skills’. Ha-ha-ha! You want ‘examples of code’, you’ll get them. :) The project will be on @work-with-data organization, I will fork it to alxfed, then request pullbacks, etc. etc. just to demonstrate my ‘command of the version control’ hehe. The lab journal of the project will be in this post on my github blog.

  • 9:00 am created https://github.com/work-with-data/time-to-test-pytorch1 , added .idea to .gitignore;
  • 9:05 am forked it to https://github.com/alxfed/time-to-test-pytorch1 ;
  • 11:30 posted some code of a standard iris test, because it works;
  • 12:10 made a pull request, approved the merge in @work-with-data.

    Sunday, January 27,
  • 8:30 pm found the iris dataset on Kaggle, modified the code slightly, uploaded it to a kernel.
  • 9:22 pm the kernell runned fine (5+ sec execution time), make it public there.

…to be continued.