BNS MDC bug fixes

[wbenoit26]

@deepchatterjeeligo @bhgupta20 This should be enough to get things running. The biggest issue was an API change in the Decimator object, which we should have made a minor release for, rather than a point release. AMPLFI uses v0.7.8, while the preprocessor was written expecting at least v0.7.10. As a workaround, I've pinned the version in online to v0.7.8 and modified the preprocessor accordingly.

Otherwise, there were just some minor corrections.

[bhgupta20]

@wbenoit26 thank you for the fixes. I have a question, why did you change this to 12 per day?

Also, I think split is not an argument for the forward function, but instantiated in the beginning, I think that is why the test is failing?

[wbenoit26]

I've reverting the ml4gw version back to when split was an argument in the forward function.

I changed the threshold to a FAR of 12 per day because that's the standard threshold for uploading an event during the MDC. We just had the default set incorrectly.

[deepchatterjeeligo]

@wbenoit26 I am wondering if it is a better idea to put the requires-python >= 3.10 and then updating the lock files as a separate MR to main, and rebasing this branch. Those changes seem atomic enough. You can create a new branch from main and cherry-pick the pyproject.toml changes.

[wbenoit26]

@deepchatterjeeligo I was thinking the same thing. Once I get the environment issues sorted out, I'll make the changes to main.

[wbenoit26]

@deepchatterjeeligo There's actually enough differences between what I'd want to do on main and what I need to do here that I think it makes more sense to do things separately. Eventually, we can update AMPLFI to use to same Decimator API and train a new model, etc. at which point the environments will be back in sync.