why add loss_norm in loss

[chenzyhust]

After loss_norm, all loss items are 1.0, is this a bug?

[zehuichen123]

Hi, I have the same issue.

2023-03-13 14:56:05,325 - mmdet - INFO - Epoch [1][50/3517]     lr: 7.973e-05, eta: 2 days, 3:53:00, time: 2.214, data_time: 0.181, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.0185
2023-03-13 14:57:48,944 - mmdet - INFO - Epoch [1][100/3517]    lr: 9.307e-05, eta: 2 days, 2:11:33, time: 2.072, data_time: 0.047, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.2678
2023-03-13 14:59:32,889 - mmdet - INFO - Epoch [1][150/3517]    lr: 1.064e-04, eta: 2 days, 1:39:35, time: 2.079, data_time: 0.048, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.6240
2023-03-13 15:01:15,588 - mmdet - INFO - Epoch [1][200/3517]    lr: 1.197e-04, eta: 2 days, 1:14:03, time: 2.054, data_time: 0.047, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.8502
2023-03-13 15:02:58,924 - mmdet - INFO - Epoch [1][250/3517]    lr: 1.331e-04, eta: 2 days, 1:01:37, time: 2.067, data_time: 0.047, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.9961
2023-03-13 15:04:42,935 - mmdet - INFO - Epoch [1][300/3517]    lr: 1.464e-04, eta: 2 days, 0:55:54, time: 2.080, data_time: 0.053, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.9350
2023-03-13 15:06:26,404 - mmdet - INFO - Epoch [1][350/3517]    lr: 1.597e-04, eta: 2 days, 0:49:09, time: 2.069, data_time: 0.045, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.6893
2023-03-13 15:08:09,362 - mmdet - INFO - Epoch [1][400/3517]    lr: 1.731e-04, eta: 2 days, 0:41:52, time: 2.059, data_time: 0.046, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.4040
2023-03-13 15:09:53,657 - mmdet - INFO - Epoch [1][450/3517]    lr: 1.864e-04, eta: 2 days, 0:39:58, time: 2.086, data_time: 0.045, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.3266
2023-03-13 15:11:37,994 - mmdet - INFO - Epoch [1][500/3517]    lr: 1.997e-04, eta: 2 days, 0:38:14, time: 2.087, data_time: 0.048, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.1977
2023-03-13 15:13:21,793 - mmdet - INFO - Epoch [1][550/3517]    lr: 2.000e-04, eta: 2 days, 0:35:08, time: 2.076, data_time: 0.044, memory: 13109, loss_depth: 1.0000, loss_voxel_ce_c_0: 1.0000, loss_voxel_sem_scal_c_0: 1.0000, loss_voxel_geo_scal_c_0: 1.0000, loss_voxel_lovasz_c_0: 1.0000, loss: 5.0000, grad_norm: 3.0625

[JeffWang987]

please see https://github.com/JeffWang987/OpenOccupancy/issues/4