Create the CIs for Icy and deepImageJ using JDLL

[carlosuc3m]

No description provided.

[carlosuc3m]

Hello @FynnBe this is the CI for deepimagej, exclusively.
I am now doing two separate CIs for deepimagej and deepicy testing the actual software instead of just the backend (JDLL). In my opinion, now the CI mimics the reality better.

Please, if you can, have a look at the generated reports to see if they are what it is needed:
https://github.com/carlosuc3m/collection/actions/runs/13031819471
https://github.com/carlosuc3m/collection/actions/runs/13031819471/artifacts/2504412013

As expected, dij only has issues with pytorch_state_dict models.

The only thing remaining to do is to check why are there some models that produce an output different to the expected one. Once that is fixed it should be ready to be merged. Then I can do another pull request for the Icy CI, or would you rather have both of them together?

This CI only checks for published models, should it check draft models too?

REgards,
Carlos

[FynnBe]

Hello @FynnBe this is the CI for deepimagej, exclusively. I am now doing two separate CIs for deepimagej and deepicy testing the actual software instead of just the backend (JDLL). In my opinion, now the CI mimics the reality better.

awesome 👍

Please, if you can, have a look at the generated reports to see if they are what it is needed: https://github.com/carlosuc3m/collection/actions/runs/13031819471 https://github.com/carlosuc3m/collection/actions/runs/13031819471/artifacts/2504412013

looks great! 👍

The only thing remaining to do is to check why are there some models that produce an output different to the expected one. Once that is fixed it should be ready to be merged.

yeah.. I'm currently working on an update to spec to set more generous default tolerances. This is probably something you'd want to check in the JDLL, too then...

Then I can do another pull request for the Icy CI, or would you rather have both of them together?

no, separate is even better 👍

This CI only checks for published models, should it check draft models too?

no, published only is good

[carlosuc3m]

Hello again @FynnBe

yeah.. I'm currently working on an update to spec to set more generous default tolerances. This is probably something you'd want to check in the JDLL, too then...

Should i just increase the tolerances to let them pass?