Hi about the dataset for FB and ConceptNet

snap-stanford / csr

Apache License 2.0

30 stars 6 forks source link

Hi about the dataset for FB and ConceptNet #1

Closed LIANGKE23 closed 1 year ago

LIANGKE23 commented 1 year ago

We find there is no data inside of the train json. Can you offer it~?

q-hwang commented 1 year ago

Hi! They are left empty intentionally to represent the lack of a curated meta-training set. Instead, we have the pretrain split that contains all the rest triplets uncurated, and our code uses this pretrain mode to learn over it.