text generation pytorch causal-lm

gpt-neo-1.3B Korean version

PPL on Oscar Korean text dataset = 46.0