Two topics (after some in-person discussion)

• Looper tagger that only excludes sigma region around each cluster, instead of full rectangular window
• Create tpc-native-clusters file after e.g. NN application and perform tracking QA

Only minor updates

• Looper tagging: Seems to be working, but somehow I made my code super slow
• tpc-native-clusters: Writing done, format should be compatible, but when comparing: new file is much smaller than the original file -> Need to dig a bit

To-Do list for the end-of-yer break

• Retrain 3D classification with only fully-connected layers
• Train 3D regression
• Train 2D regression with one output per network and 2 networks
• Perform tracking QA: To be seen how easy it is or if it takes more time (but at least want to get started)
• Try inclusion of Pytorch as alidist recipe
• Modify QA task: Potentially complete rewrite to organize a bit more and find / avoid bugs, write results to unified QA.root file, etc.
• Get back to PyTorch about Conv3D layer speed