Can you show the configuration for the JEM baseline in Table 1?

[sndnyang]

Hi, Hao

Good paper and good results. I have a question about the baseline of JEM in table 1. How to achieve this result ~94.4% with JEM? In table 4, it's just ~92.9% which is reproducible.

I tried something like enabling the batch norm. But it doesn't work?

So, can you show the configuration for the JEM baseline in Table 1?

Thank you so much!