metadata
license: mit
Datasets:
Pile 10B (from NeelNanda/pile-tokenized-10b) OpenWeb 10B (from karpathy nanoGPT)
Tokenizer used:
EleutherAI/gpt-neox-20b
Data type:
new_data_20b.bin (20B tokens) val.bin (4M tokens)
Feel Free to use it as you please!
-just follow the rules of ofiginal datasets :)