Closed dujiangsu closed 2 years ago
As in the title, it is not essential to take two dirs for gpt-2 and -3 respectively. Besides, train_gpt.py is for transformers.GPT2LMHeadModel, not the Colossal version.
Hi @Gy-Lu , can you take care of this issue as well?
Fixed by #28
As in the title, it is not essential to take two dirs for gpt-2 and -3 respectively. Besides, train_gpt.py is for transformers.GPT2LMHeadModel, not the Colossal version.