Conversation
after backbone/head split the model output is not longer a dict for regression task
TjarkMiener
requested review from
BastienLacave,
Pablitinho,
nietootein and
rcervinoucm
This script can be used to repair existing files, in case only one task is not valid
fix also sky direction reco for stereeo docstring polishing
code was breaking since the feature vector was never set to None
nietootein
left a comment
nietootein
left a comment
There was a problem hiding this comment.
Hi, @TjarkMiener . In my understanding this PR is already ready to review and has my go. Thanks so much for catching that bug and all the improvements that accompany this PR.
removing the env for the IT cluster
fixes also the rotation from the new dl1dh version to be relesased v0.14.1
|
Hi @nietootein, thanks! I'm almost done. Performing the last checks. I will open a bug fix PR on dl1dh, since the pixel rotation was not properly propagated. I think we need to release then a bug fix release of the dl1dh v0.14.1. Then, we need to update the dl1dh version here in the env and pyproject, before merging this PR and releasing ctlearn v0.10.0. I will try to do it by tomorrow noon. Can you please check here (and also in the dl1dh) if the codemeta and zenodo needs any changes? Thanks! |
code checks if prefix_telescopes col has not the bool as dtype and then apply the fix. If it will be fix later in ctapipe, this code will not break.
|
Once dl1dh v0.14.1 is released we'd need to re-run the CI here, which will then pass @nietootein |
|
Thanks so much, @TjarkMiener! |
2 participants
There are some bug fixes needed for the prediction tools. I'm not 100% done yet, so I will open it as a draft.