Wrong epoch count on training (Bug?)

[AfterHAL]

Hi.
I didn't pay attention to it before, but I am currently testing a training base on the "Sana_600M_img512.yaml" config file with a dataset ok 20k images (single 4090 GPU / batch 4), and the reported epoch after 80000 steps is still epoch 1:

2024-12-05 21:55:33 - [Sana] - INFO - Epoch: 1 | Global Step: 81780 | Local Step: 4780 // .......

(And the reported time at the begining is not the actual time of the system)

[Deng-Xian-Sheng]

You should have 4 epochs, and then your second epoch is at 40000