Sorry. I did resolve it, but have no memory of how. I’m up to using transformers 4.9.2 now and do not have the issue and do not need to make changes to their run_classification script. (I’m using the pytorch version).
1 Like