In order to determine randomness causing different outputs between notebook runs, I went through the ReviewNB and looked at all the diffs between two runs.
Identified differences:
pd.DataFrame.sample randomly samples rows from a data frame. Setting random_state fixes this.
A raster plot on ecephys_sessions randomly samples colors.
tmp directories are assigned a UUID
printouts of transfer speeds vary when files are reached
In order to determine randomness causing different outputs between notebook runs, I went through the ReviewNB and looked at all the diffs between two runs.
Identified differences:
random_state
fixes this.Only (1) was fixed in this PR.