BiasBios notebook fails - dimensionality of data mismatch

[efemeryds]

Dear Shauli, for my Master research thesis purposes I am trying to re-create the biasbios_bert.ipynb notebook.

• I set the virtual environment to use Python 3.7
• I downloaded the data using download_data.sh.
• When I run the notebook I get the error that the lengths of the text data and encoding data mismatch

I tried using different libraries versions thinking that maybe the reading of the pickles fails, but I keep getting the same error.

I also tried running the run_bias_bios.sh locally to get the encoding from the text hoping that the dimensions would match then, but it is too computationally expensive for my current resources and I could not finish the training.

Was the data updated since the notebook was run the last time? And if the data is correct, do you have any other ideas why I cannot re-create the notebook?

Best regards, Alicja

[afsharim]

I have the same problem.