Upgrade torch to 1.11.0

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

Upgrade torch to 1.11.0 #19

Closed A669015 closed 2 years ago

A669015 commented 2 years ago

This PR upgrades the torch version used to 1.11.0 and associated it the 11.3 cuda version by default in the requirements.

Additionally it:

upgrades the torch-sparse version to be compatible with the default cuda version supported;
fixes the nox installation on CI/CD pipeline (the integrated action was still using the latest python version whatever the python version used in the matrix),
fixes the cuda/torch combination when cuda in not available on the system (that is the case on Github CI/CD),
add nox session verbosity to ON, for easier debug purpose.

github-actions[bot] commented 2 years ago

Coverage report for weather_forecast/gwd/ use-case

File	Coverage		Missing
All files	`94%`	:white_check_mark:
data.py	`98%`	:white_check_mark:	`157`
models.py	`89%`	:white_check_mark:	`37-40` `90` `99`
trainer.py	`88%`	:white_check_mark:	`75-76` `80`

Minimum allowed coverage is 80%

Generated by :monkey: cobertura-action against be0a45cac66c8e5a272d44e7fa502b04decc46fb

github-actions[bot] commented 2 years ago

Coverage report for combustion/unets/ use-case

File	Coverage		Missing
All files	`91%`	:white_check_mark:
data.py	`91%`	:white_check_mark:	`59` `70` `72` `113` `123` `129`
models.py	`87%`	:white_check_mark:	`41-42` `45` `48`
trainer.py	`75%`	:x:	`59-62` `66-67` `71`
unet.py	`96%`	:white_check_mark:	`83` `89`

Minimum allowed coverage is 80%

Generated by :monkey: cobertura-action against be0a45cac66c8e5a272d44e7fa502b04decc46fb

github-actions[bot] commented 2 years ago

Coverage report for combustion/gnns/ use-case

File	Coverage		Missing
All files	`80%`	:white_check_mark:
config.py	`68%`	:x:	`74-77` `81` `85-90` `94-101`
data.py	`81%`	:white_check_mark:	`110-137` `163` `217`
models.py	`68%`	:x:	`85` `116-131` `156-160` `210-214` `234-238`
plotters.py	`98%`	:white_check_mark:	`55-56`
trainer.py	`76%`	:x:	`70-75` `79-80`

Minimum allowed coverage is 80%

Generated by :monkey: cobertura-action against be0a45cac66c8e5a272d44e7fa502b04decc46fb