Missing dev split

[senisioi]

Trying to replicate your results, I stumbled over a missing dataset ../data/dev_100M_tokenized.bin
Do you recommend splitting the hf dataset to get the dev?

Great work!

[davda54]

Hi, we used the official BabyLM validation split :)