Too much validation time #228
Description
Is it normal for the validation time of an epoch to be longer than the training time of an epoch?
Train | BoxLoss | DFLLoss | BCELoss | ━━━━━━━━━━━━━━━━━━━━━━━━━ 0/249 5:37:27 • 23 days, 0.00it/s Batch | 3.23 | 2.70 | 3.48 | ━━━━━━━━━━━━━━━━━━━ 35947/35947 2:32:13 • 0:00:00 4.33it/s Valid | mAP : 1.26 | mAP50 : 8.17 | ━━━━━━━━━━━━━━━━╸━━━━━━━━ 3006/4494 3:05:12 • 3:03:17 0.14it/s
I'm aware that my dataset is large (575139 train images and 71893 valid images), but I've trained other models with the same dataset and this seems odd to me.
How can I improve valid speed?