Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[WIP] Switch to poetry for package management #67

Draft
wants to merge 8 commits into
base: master
Choose a base branch
from
Draft

Conversation

lukehsiao
Copy link
Owner

@lukehsiao lukehsiao commented Mar 23, 2021

Poetry has gained a lot of popularity, and one key feature is the lockfile. If we include that we can be more effectively capture the exact versions and dependencies needed to run the pipeline.

Before this is merged:

  • The pipeline should be re-run to make sure everything works as expected
  • The README instructions are all up to date with the new commands

Added wheel to the dev requirements.txt to avoid the missing Cython
error that occurs when building from source [1]. This also removed the
need for cython as an extra dependency of HACK.

Numpy is moved up in the dependency list so that it is present before
scikit-learn is attempted to be installed, which will fail otherwise.

Postgresql isn't specifically installed since it is included in the
GitHub Action VM.

[1]: scikit-learn/scikit-learn#19068
This also updates the dependencies to the latest compatible versions and
fixes all the lint errors that result.
This way, the CLI scripts are automatically installed when running
`poetry install` and are available for use as described in the README.
@lukehsiao
Copy link
Owner Author

I'm starting another full run of the transistor dataset now to verify this switch doesn't break anything.

Snorkel is restricting the networkx dependency to <2.4, which doesn't
support Python 3.9. Consequently, we can't use 3.9 either.
@lukehsiao
Copy link
Owner Author

lukehsiao commented Apr 14, 2021

Here is one e2e log with this poetry setup:
2021-03-31_transistors.log

The intermediate numbers are slightly different, but the end results are very similar. I'll run the opamps and circular connectors next.

@lukehsiao
Copy link
Owner Author

lukehsiao commented Apr 21, 2021

Here's an e2e log for opamps:

2021-04-14_opamps.log

This shows a significant regression on the gain relation that should be looked at more closely.

@lukehsiao
Copy link
Owner Author

Here's an e2e log for circular connectors:

2021-04-21_circular-connectors.log

This also shows a significant regression in F1 (~5 pts)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant