Inconsistency from B-Train found in ~/cernbox/hysteresis/dipole/datasets/train/SPS-BTRAIN-20240325-163646---20240325-164243_bc_awake.parquet where one field on the AWAKE cycle is flat on flat bottom.
~/cernbox/hysteresis/dipole/datasets/train/SPS-BTRAIN-20240312-142900---20240312-143327_bc_sftpro_md_lhc.parqueton an LHC cycle~/cernbox/hysteresis/dipole/datasets/train/SPS-BTRAIN-20240312-151639---20240312-152432_bc_sftpro_lhc_scrubbing.parqueton scrubbing supercycle (LHCMD2)
Training and validation datasets in Dipole datasets has been adjusted on CERNBox.
2024-11-21
We notice significant drift cycle-to-cycle from data measured 20241105
This is due to inconsistent (or lacking) marking on the MD cycle used, as well as the ion cycle, so the field was not being marked at all, and hence continuously drifting.
