Misc fixes on combustion/gnns

AI4SIM / model-collection

This project contains a collection of deep learning models developed by the AI4Sim team with various partners. This is structured on the basis of use-cases providing canonical PyTorch Lightning pipelines allowing to train neural network models that are able to surrogate various physical processes.

Apache License 2.0

16 stars 5 forks source link

Misc fixes on combustion/gnns #11

Closed A669015 closed 2 years ago

A669015 commented 2 years ago

This PR proposes several fixes identified by running the combustion/gnns/trainer.py on toy dataset with a full cpu machine:

complete the requirements.txt
add missing parameters in the configuration files
remove the _/ folder
allow the dataloader to accept different dataset size

github-actions[bot] commented 2 years ago

:white_check_mark: Result of Pytest Coverage

---------- coverage: platform linux, python 3.9.10-final-0 -----------	Name	Stmts	Miss
combustion/gnns/init.py	0	0	100%
combustion/gnns/config.py	21	21	0%
combustion/gnns/data.py	71	71	0%
combustion/gnns/models.py	82	82	0%
combustion/gnns/plotters.py	103	103	0%
combustion/gnns/test.py	48	48	0%
combustion/gnns/tests/test_snippet.py	12	1	92%
combustion/gnns/trainer.py	26	26	0%
TOTAL	363	352	3%

~ 2 passed in 0.10s ~

rmslm commented 2 years ago

FYI : also we've decided a while ago to squash our commits when asking for PR if you could do so that would be great thx :)

A669015 commented 2 years ago

Not sure to see your point ... in the history there is a lot of commit that are not squashed. You mean when the PR will be merged, I should choose the "squash and merge" strategy?

rmslm commented 2 years ago

yes that's right, in the history all are not squashed my bad.... Nonetheless, that what was decided.