Can not reproduce classification accuracy for efficientvit-B1 and B2 on Imagenet.

[xiaoshingshing]

Hi, I use the default setting and the Single-Node Training Examples command line at here, but cannot reproduce the classification accuracy.

For efficientvit-b1-224, I reach 79.26% top1 accuracy while you report 79.39%.

For efficientvit-b2-224, I reach 81.85% top1 accuracy while you report 82.10%.

Are there any suggestions to reproduce the reported results?