How to edit encoder.py to work with new release of Grover 1.5B Mega model?

[GenTxt]

Thanks for this repo. Have it sampling on an 8 gig machine using the 1.5B OpenGPT-2 model. It generates interesting and easy to read text based on prompts. None of that .jsonl file mess to deal with.

Your revised "contextual_generate_cli.py" and "encoder.py" work perfectly with the above model but when I attempt to use these files with the recently released Grover 1.5B Mega model it generates nothing but garbage:

https://github.com/rowanz/grover

friend hepr Inox liter/ H friend investments 1 doWikic it he interestal Wil Alternative-ic it he interest m g***ess mrian 1 do standards postighation Google Policy der Rosenber

I see that both files are from the original Grover repo and heavily commented-out with a few new code editions to accept text input and generate text output. I've tried several edits but it either results in coding errors or generates the same garbage output above.

Any suggestions on how these files can be modified and used with the Grover Mega model?

Thanks