Closed ikowalec closed 7 months ago
The CI tests will likely fail due to missing imports. Rebuilding torch
as part of CI is not an option due to size - @GabrielBram ideas?
Edit: for now using a dry_run
flag and EMT()
calculator
Attention: 19 lines
in your changes are missing coverage. Please review.
Comparison is base (
68fddf9
) 86.58% compared to head (672132b
) 87.50%.
Files | Patch % | Lines |
---|---|---|
carmm/run/workflows/react_mace.py | 87.39% | 15 Missing :warning: |
carmm/run/workflows/helper.py | 71.42% | 4 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Look through the MACE ASE calculator, they unfortunately import torch at the head so I don't believe there's a way around this. Would this require Andy to update the CI test runner?
Look through the MACE ASE calculator, they unfortunately import torch at the head so I don't believe there's a way around this. Would this require Andy to update the CI test runner?
We would have to cache the pytorch installation somehow - we can't just download and rebuild 2.5gb of packages on each pull request. Something to think about.
Edit: Apparently there is a cached version of pytorch, so dependancies are building only ~60 sec longer with torch, torchvision and mace in the CI test
Look through the MACE ASE calculator, they unfortunately import torch at the head so I don't believe there's a way around this. Would this require Andy to update the CI test runner?
now resolved by amending the main.yml
Now removed the [WIP] tag.
TODOs:
react.py
andreact_mace.py
Done: