More training data

[renau]

Nice work,

One suggestion, why not to use the already quite formatted (not exactly but close) data from rosseta code to populate. There are not so many examples even in the 20K.json, This will significantly improve the amount of code to train on (and languages).

E.g: look at this: https://rosettacode.org/wiki/Accumulator_factory

[sahil280114]

I just didn’t look at that option, if you’d be able to contribute the formatted data as a PR I’ll be happy to redo the training run

[iraqigeek]

Found about Rosetta code a few minutes ago reading this issue. After a couple of minutes of googling, I found this repo on 🤗:
https://huggingface.co/datasets/bigcode/rosetta-code