OS: Windows 10 build 18363
DeepLabCut Version: 2.2b5
Anaconda env used: dlc-windowsGPU
Following these and these instructions, I am trying to convert old labels to multianimal. While the "multianimalbodyparts" are correctly saved in the new .h5 file, the "uniquebodyparts" are dropped by the function "convert2_maDLC" (or perhaps the config.yaml is not correctly prepared?).
config.yaml reads as follows:

Calib1 to Calib6 are not present in the new CollectedData[...].h5 file. I went through the code but could not find any issue.
Here a "check_labels" output (the six calibration markers are on the top-left corner of the snapshot):

Would you have any hint? Thank you guys!
Hi! If you pull from source this behavior is changed, ie unique will stay - we will update to 2.2b6 today (the docs got a little too far ahead; before I said leave unique blank - apologies!)
Thank you Mackenzie, helpful as usual! Will let you know how it works for me.
Thanks for testing the master, will be released to pypi soon (let us know if there is a problem).
Wonderful, thanks Alex!
"uniquebodyparts" are not dropped any more with DLC 2.2b6 when converting labels from single-to multianimal. Thank you!
Most helpful comment
Thank you Mackenzie, helpful as usual! Will let you know how it works for me.