kmaninis
commented
5 years ago It probably has to do with your image size. We do not divide the loss by the number of elements in the prediction, thus the high number. Is your image size comparable to ours (480p)?
I come across a problem and very confused:
2019-02-23 22:36:17.311281 Iter 586: Training Loss = 79458.1719 2019-02-23 22:36:19.815631 Iter 588: Training Loss = 37970.0039 2019-02-23 22:36:22.456881 Iter 590: Training Loss = 46907.9766 2019-02-23 22:36:24.121243 Iter 592: Training Loss = 21860.0176 2019-02-23 22:36:25.718174 Iter 594: Training Loss = 49590.5156 2019-02-23 22:36:27.451493 Iter 596: Training Loss = 90751.0938 2019-02-23 22:36:29.451918 Iter 598: Training Loss = 17363.9922 2019-02-23 22:36:31.202461 Iter 600: Training Loss = 27310.3105 2019-02-23 22:36:33.080727 Iter 602: Training Loss = 81702.3984 2019-02-23 22:36:35.805301 Iter 604: Training Loss = 43151.9375 2019-02-23 22:36:38.184319 Iter 606: Training Loss = 26587.7422 2019-02-23 22:36:40.321566 Iter 608: Training Loss = 214049.2031 2019-02-23 22:36:42.982664 Iter 610: Training Loss = 11704.9385 2019-02-23 22:36:44.731441 Iter 612: Training Loss = 173234.9844 2019-02-23 22:36:46.219128 Iter 614: Training Loss = 1401.4938 2019-02-23 22:36:47.852391 Iter 616: Training Loss = 52776.4648 2019-02-23 22:36:49.122615 Iter 618: Training Loss = 31279.1348 2019-02-23 22:36:52.050488 Iter 620: Training Loss = 12059.2002 2019-02-23 22:36:53.660343 Iter 622: Training Loss = 67013.9453 2019-02-23 22:36:55.264316 Iter 624: Training Loss = 12743.4521 2019-02-23 22:36:57.738322 Iter 626: Training Loss = 10916.0781 2019-02-23 22:36:59.682404 Iter 628: Training Loss = 66693.2812
can you tell me why my loss is so high?can you tell me your loss in your training~ thank you so much!!!