Can you include code for this to handle GPT-Neo-125M

[thistleknot]

I was considering diving into the source to do it, but I don't see any clear indications of huggingface models. I assume you must be downloading the weights and then applying them to an empty pytorch model.

[thistleknot]

I'm actually trying to borrow a prompt engineering method used for simple-gpt2
https://github.com/RossSong/GPT2-Question-Answering/blob/master/QA.ipynb
for retooling with happytransformer.

EricFillion/happy-transformer#310