two training loss per step

[fzuo1230]

hello, i was training wan2.1-14b-i2v-720p, but i found that i get two loss print after one step:

Average Epoch time: 231.10 seconds
Average Peak memory 68550.57MiB

Epoch 0: 1%|▏ | 3/217 [13:05<15:33:34, 0.00it/s, v_num=0, train_loss=0.187]
Epoch 0: 1%|▏ | 3/217 [13:05<15:33:34, 0.00it/s, v_num=0, train_loss=0.118]

Is this normal or am I doing something wrong?