jdjin3000 / PRG-MoE

6 stars 2 forks source link

reproduce the results of the paper #5

Open yangweiboyang opened 1 year ago

yangweiboyang commented 1 year ago

Hello,It's very glad to read your paper, thanks for your contribution.

I have run your code yesterday, and I don't know why i don't get the result as your paper shows。

this is my result, I just changed the batch size from 5 to 4, and use one gpu. image

and this is your paper's result image

is there any other setting? Looking forward to your reply.