LiuMinghui

Results 2 comments of LiuMinghui

Thanks to the author for contributing the code! When I run **src_pretrain** directly, both the **loss and feature** become **‘nan’** on the second batch. Do you have this problem when...

Thank you very much!