nora1827
commented
1 year ago The training log is as follows: 2022/12/28 13:05:30 epoch: 1, lr: 0.0000100, loss: 0.0000, p1: 44.91, p2: 36.12 2022/12/28 15:53:41 epoch: 2, lr: 0.0000095, loss: 0.0000, p1: 45.59, p2: 36.28 2022/12/28 18:42:36 epoch: 3, lr: 0.0000090, loss: 0.0000, p1: 45.15, p2: 36.28 2022/12/28 21:29:36 epoch: 4, lr: 0.0000086, loss: 0.0000, p1: 44.73, p2: 35.97 2022/12/29 00:15:49 epoch: 5, lr: 0.0000081, loss: 0.0000, p1: 45.02, p2: 36.22 2022/12/29 03:02:09 epoch: 6, lr: 0.0000041, loss: 0.0000, p1: 45.19, p2: 36.40 2022/12/29 05:49:23 epoch: 7, lr: 0.0000039, loss: 0.0000, p1: 44.78, p2: 36.04 2022/12/29 08:36:25 epoch: 8, lr: 0.0000037, loss: 0.0000, p1: 45.25, p2: 36.30 2022/12/29 11:23:41 epoch: 9, lr: 0.0000035, loss: 0.0000, p1: 44.82, p2: 36.02 2022/12/29 14:10:43 epoch: 10, lr: 0.0000033, loss: 0.0000, p1: 45.17, p2: 36.26 2022/12/29 16:57:16 epoch: 11, lr: 0.0000017, loss: 0.0000, p1: 44.90, p2: 36.08 2022/12/29 19:46:24 epoch: 12, lr: 0.0000016, loss: 0.0000, p1: 45.02, p2: 36.19 2022/12/29 22:35:25 epoch: 13, lr: 0.0000015, loss: 0.0000, p1: 44.99, p2: 36.12 2022/12/30 01:24:01 epoch: 14, lr: 0.0000014, loss: 0.0000, p1: 45.55, p2: 36.22 2022/12/30 04:09:52 epoch: 15, lr: 0.0000014, loss: 0.0000, p1: 44.94, p2: 36.02 2022/12/30 06:57:28 epoch: 16, lr: 0.0000007, loss: 0.0000, p1: 45.01, p2: 36.15 2022/12/30 09:45:29 epoch: 17, lr: 0.0000006, loss: 0.0000, p1: 45.03, p2: 36.17 2022/12/30 12:32:58 epoch: 18, lr: 0.0000006, loss: 0.0000, p1: 45.14, p2: 36.15 2022/12/30 15:20:35 epoch: 19, lr: 0.0000006, loss: 0.0000, p1: 45.30, p2: 36.41 2022/12/30 18:09:03 epoch: 20, lr: 0.0000006, loss: 0.0000, p1: 45.19, p2: 36.27
Vegetebird
When try to python main.py --refine --lr 1e-5 --reload --previous_dir [your model saved path], the loss is always 0 overall process, is it normal? I can't get the same result in the paper 《Exploiting Temporal Contexts with Strided Transformer for 3D Human Pose Estimation》.