Add wet run testing to github

[mackenziesnyder]

MSE results:

• run 1: 1.0965
• run 2: 1.0965

To do:

• decide on acceptable MSE, right now I've set it to 1.2mm (all 32 afids, x, y, z included).

Note:

• wet runs take around 20 minutes to run due to model downloads
• just for python3.10

for issue: #41

[Dhananjhay]

Superb job @mackenziesnyder! I was wondering if it's possible to cache the CNN models so that we don't have to download them everytime a wet-run test is executed?

[Dhananjhay]

We can reproduce the MSE to the exact decimal, i.e., 1.0965 on every run, so I was wondering if we need a threshold to begin with. What do you guys think? @mackenziesnyder @ataha24. Wet-run testing is basically to run the pipeline and make sure the rules are running properly and we can get a .fscv file out in the end.

[mackenziesnyder]

Superb job @mackenziesnyder! I was wondering if it's possible to cache the CNN models so that we don't have to download them everytime a wet-run test is executed?

i will look into this! that would make the run a lot faster

[mackenziesnyder]

We can reproduce the MSE to the exact decimal, i.e., 1.0965 on every run, so I was wondering if we need a threshold to begin with. What do you guys think? @mackenziesnyder @ataha24. Wet-run testing is basically to run the pipeline and make sure the rules are running properly and we can get a .fscv file out in the end.

I can set it to this exact value!

[ataha24]

Sounds good to me!

[mackenziesnyder]

@Dhananjhay @ataha24 I've made the wet runs to be triggered on pull request, added --fidqc and --stereotaxy STN to the wet run, and removed the code that had attempted to cache the model dir. This one should be good to go!

[Dhananjhay]

This looks awesome, @mackenziesnyder! One small thing - I just merged PR #46 where I removed the logic of running dry run tests using poethepoet and I was wondering if we could achieve the same here. You can refer to the main branch for guidance. Thanks again for the amazing work!

[mackenziesnyder]

This looks awesome, @mackenziesnyder! One small thing - I just merged PR #46 where I removed the logic of running dry run tests using poethepoet and I was wondering if we could achieve the same here. You can refer to the main branch for guidance. Thanks again for the amazing work!

Will do!! I will refactor it to match !

[mackenziesnyder]

@Dhananjhay I have tested this out and it ran successfully, I had to add an additional flag --conda-frontend conda because of a mamba dependency with conda. I think this is specific to the fact we do the dry and wet runs with poetry

[Dhananjhay]

Amazing work @mackenziesnyder! I'll review this PR later today and merge it in, thank you!

[mackenziesnyder]

Amazing work @mackenziesnyder! I'll review this PR later today and merge it in, thank you!

thanks!

[Dhananjhay]

• Replaced the Poetry-based environment with a Conda-based environment (run the pipeline by calling the run.py executable).
• Implemented caching for Conda environments to reduce setup time.
• Cached CNN models to avoid repeated downloads.
• Resulted in optimizing the wet run testing timeline by ~ 58%.
• Fixed some dry run tests by defining number of cores, i.e., c1 flag.

[Dhananjhay]

I think this is good to merge. Let me know once you've reviewed the changes @mackenziesnyder!

[mackenziesnyder]

do we still need the --conda-frontend mamba here? I added this because it was needed when originally using poetry but when I activated a conda env and ran autoafids during a test on cbs I no longer needed this flag @Dhananjhay

[Dhananjhay]

Yeah, it's a little tricky to run with poetry but mainly we use it here to speed up the pipeline. When using conda, Snakemake by default injects --conda-frontend conda to resolve the dependencies; we instead use --conda-frontend mamba which is much faster at that!

[mackenziesnyder]

Ok I see, that makes sense! thanks, the code look good!