Updates for January 24

[bfhealy]

General announcements

• New training data on YSOs and CVs are being uploaded to the golden dataset on Fritz (group 1458).

Pull requests

• Fix callbacks for pre-trained model, use legacy optimizer #214
• Save obj_id for uploaded sources #215
• Rework uploading of specific classifications #216
• Toggle upvoting for new classifications #217
• Checkpoint obj_id saves #218
• Handle missing photometry during source upload #220
• Add cone search for ZTF IDs #222
• Add more flexibility to get_features.py #223

Issues

• Option to delete existing classifications when posting updated ones #212
• Uploading training sources without good photometry #219
• Include filter ID in feature catalogs #221
• Handling multiple ML label probabilities for the same source #224

Going forward

Progress continues on retraining the classifiers and posting a new set of ML classifications. The plan is for this new active learning set to be available by the end of next week. The remaining tasks are:

• Rework training set downloads to include all available rows of features
• Merge new feature imputation code (Consistent/custom imputation strategy #207)
• Rerun training with adjusted hyperparameters
• Upload active learning sample to Fritz, accounting for multiple classification probabilities from different feature rows