pytorch/multipy

Create a Contributing.md

PaliC opened this issue · 0 comments

PaliC commented

TL;DR: Build torch::deploy in OSS and run unit tests + document what you did.

Our Contributing.md was forgotten and is at the point where it "exists". We have gotten some complaints from developers that our development workflow instructions are not great. Fortunately, this presents a pretty good learning opportunity for newer developers.

  • Get on a Debian based system as that will make things easier.
  • Clone the repo using a standard git workflow
  • Follow the instructions for the Pip Installation up until actually running pip install then run these instructions to build
  • Document this build process in CONTRIBUTING.md
  • Run our unit tests
  • Document this build process in our CONTRIBUTING.md
  • Make some temporary changes into multipy/runtime/deploy.cpp (these can be print statements/ asserts/ whatever just don't commit them).
  • Test to confirm that your changes have occurred, and document the process in CONTRIBUTING.md.
  • Make some temporary changes into multipy/runtime/interpreter/interpreter_impl.cpp (these can be print statements/ asserts/ whatever just don't commit them).
  • Test to confirm that your changes have occurred, and document the process in CONTRIBUTING.md and mention that this process is only needed for changes that are specific to individual interpreters.
  • Create a Pull Request and press the button on the PR to import it into FBCode. The only file that is changed should be CONTRIBUTING.md, and submit it for review.

I expect there should be some trial and error with some of these steps. When you are recording things into the CONTRIBUTING.md think about how a user would avoid those situations. If you get stuck anywhere feel free to reach out to Sahan Paliskara.