EthanC111 opened this issue a year ago · 0 comments
We definitely need to continue pretraining our base model on a large number of math-related corpus.