Datasetpublic

zuom/P3-unified

Currently contains the 35 datasets used for the T0 Held-in training. All tokenized inputs and targets are removed, and only the prompt/response are stored.

sourceHugging Faceupdated 1y agoView on Hugging Face

0likes747downloads

Dataset Card

Currently contains the 35 datasets used for the T0 Held-in training. All tokenized inputs and targets are removed, and only the prompt/response are stored.