decode.loss_seg and mix.decode.loss_seg

[jiuyuzhang]

Hello, I want to use my own dataset for training, but I found that the values of both decode.loss_seg and mix.decode.loss_seg drop to 0 very quickly during training, I'm not sure if the problem is with the parameters or the dataset, could you please help me to look at it, part of the training process is as follows
2023-10-26 08:41:59,207 - mmseg - INFO - workflow: [('train', 1)], max: 40000 iters
2023-10-26 08:42:44,761 - mmseg - INFO - Iter [50/40000] lr: 1.958e-06, eta: 10:04:54, time: 0.908, data_time: 0.011, memory: 9743, decode.loss_seg: 0.6010, decode.acc_seg: 13.7835, src.loss_imnet_feat_dist: 0.0329, mix.decode.loss_seg: 0.3329, mix.decode.acc_seg: 35.2100
2023-10-26 08:43:30,109 - mmseg - INFO - Iter [100/40000] lr: 3.950e-06, eta: 10:03:37, time: 0.907, data_time: 0.010, memory: 9743, decode.loss_seg: 0.5742, decode.acc_seg: 41.1594, src.loss_imnet_feat_dist: 0.0358, mix.decode.loss_seg: 0.3116, mix.decode.acc_seg: 55.3378
2023-10-26 08:44:15,362 - mmseg - INFO - Iter [150/40000] lr: 5.938e-06, eta: 10:02:17, time: 0.905, data_time: 0.010, memory: 9743, decode.loss_seg: 0.3623, decode.acc_seg: 40.3447, src.loss_imnet_feat_dist: 0.0399, mix.decode.loss_seg: 0.1926, mix.decode.acc_seg: 51.9543
2023-10-26 08:45:01,109 - mmseg - INFO - Iter [200/40000] lr: 7.920e-06, eta: 10:02:52, time: 0.915, data_time: 0.010, memory: 9743, decode.loss_seg: 0.3590, decode.acc_seg: 53.8857, src.loss_imnet_feat_dist: 0.0398, mix.decode.loss_seg: 0.1525, mix.decode.acc_seg: 66.6974
2023-10-26 08:45:47,103 - mmseg - INFO - Iter [250/40000] lr: 9.898e-06, eta: 10:03:34, time: 0.920, data_time: 0.010, memory: 9743, decode.loss_seg: 0.2208, decode.acc_seg: 56.3512, src.loss_imnet_feat_dist: 0.0395, mix.decode.loss_seg: 0.0804, mix.decode.acc_seg: 68.9543
2023-10-26 08:46:33,793 - mmseg - INFO - Iter [300/40000] lr: 1.187e-05, eta: 10:05:19, time: 0.934, data_time: 0.011, memory: 9743, decode.loss_seg: 0.1678, decode.acc_seg: 59.7599, src.loss_imnet_feat_dist: 0.0380, mix.decode.loss_seg: 0.0767, mix.decode.acc_seg: 73.5273
2023-10-26 08:47:20,687 - mmseg - INFO - Iter [350/40000] lr: 1.384e-05, eta: 10:06:44, time: 0.938, data_time: 0.011, memory: 9743, decode.loss_seg: 0.1187, decode.acc_seg: 63.4222, src.loss_imnet_feat_dist: 0.0379, mix.decode.loss_seg: 0.0478, mix.decode.acc_seg: 83.0094
2023-10-26 08:48:07,717 - mmseg - INFO - Iter [400/40000] lr: 1.580e-05, eta: 10:07:49, time: 0.941, data_time: 0.011, memory: 9743, decode.loss_seg: 0.0693, decode.acc_seg: 59.1260, src.loss_imnet_feat_dist: 0.0387, mix.decode.loss_seg: 0.0233, mix.decode.acc_seg: 85.0879
2023-10-26 08:48:54,974 - mmseg - INFO - Iter [450/40000] lr: 1.776e-05, eta: 10:08:49, time: 0.945, data_time: 0.011, memory: 9743, decode.loss_seg: 0.0573, decode.acc_seg: 62.6751, src.loss_imnet_feat_dist: 0.0399, mix.decode.loss_seg: 0.0157, mix.decode.acc_seg: 78.6375
2023-10-26 08:49:42,184 - mmseg - INFO - Iter [500/40000] lr: 1.971e-05, eta: 10:09:24, time: 0.944, data_time: 0.011, memory: 9743, decode.loss_seg: 0.0384, decode.acc_seg: 63.9637, src.loss_imnet_feat_dist: 0.0381, mix.decode.loss_seg: 0.0109, mix.decode.acc_seg: 78.2782
2023-10-26 08:50:29,605 - mmseg - INFO - Iter [550/40000] lr: 2.166e-05, eta: 10:10:00, time: 0.948, data_time: 0.011, memory: 9743, decode.loss_seg: 0.0284, decode.acc_seg: 62.9740, src.loss_imnet_feat_dist: 0.0414, mix.decode.loss_seg: 0.0087, mix.decode.acc_seg: 85.7456
2023-10-26 08:51:17,205 - mmseg - INFO - Iter [600/40000] lr: 2.360e-05, eta: 10:10:33, time: 0.952, data_time: 0.011, memory: 9743, decode.loss_seg: 0.0178, decode.acc_seg: 58.3607, src.loss_imnet_feat_dist: 0.0394, mix.decode.loss_seg: 0.0052, mix.decode.acc_seg: 79.5689
2023-10-26 08:52:04,818 - mmseg - INFO - Iter [650/40000] lr: 2.554e-05, eta: 10:10:54, time: 0.952, data_time: 0.011, memory: 9743, decode.loss_seg: 0.0158, decode.acc_seg: 57.5038, src.loss_imnet_feat_dist: 0.0420,

...

mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 82.5667
2023-10-26 09:40:46,738 - mmseg - INFO - Iter [3650/40000] lr: 5.453e-05, eta: 9:45:28, time: 0.984, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0001, decode.acc_seg: 59.8711, src.loss_imnet_feat_dist: 0.0475, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 77.3620
2023-10-26 09:41:35,947 - mmseg - INFO - Iter [3700/40000] lr: 5.445e-05, eta: 9:44:49, time: 0.984, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0001, decode.acc_seg: 71.5709, src.loss_imnet_feat_dist: 0.0414, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 86.6108
2023-10-26 09:42:25,154 - mmseg - INFO - Iter [3750/40000] lr: 5.438e-05, eta: 9:44:09, time: 0.984, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0001, decode.acc_seg: 56.7687, src.loss_imnet_feat_dist: 0.0464, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 77.5324
2023-10-26 09:43:14,629 - mmseg - INFO - Iter [3800/40000] lr: 5.430e-05, eta: 9:43:31, time: 0.990, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0001, decode.acc_seg: 59.3930, src.loss_imnet_feat_dist: 0.0530, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 80.0578
2023-10-26 09:44:04,128 - mmseg - INFO - Iter [3850/40000] lr: 5.423e-05, eta: 9:42:54, time: 0.990, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0001, decode.acc_seg: 57.2199, src.loss_imnet_feat_dist: 0.0457, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 79.8055
2023-10-26 09:44:53,500 - mmseg - INFO - Iter [3900/40000] lr: 5.415e-05, eta: 9:42:15, time: 0.987, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0000, decode.acc_seg: 53.7303, src.loss_imnet_feat_dist: 0.0461, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 76.1746
2023-10-26 09:45:42,723 - mmseg - INFO - Iter [3950/40000] lr: 5.408e-05, eta: 9:41:34, time: 0.985, data_time: 0.013, memory: 9743, decode.loss_seg: 0.0001, decode.acc_seg: 54.3883, src.loss_imnet_feat_dist: 0.0487, mix.decode.loss_seg: 0.0000, mix.decode.acc_seg: 74.0439
2023-10-26 09:47:06,376 - mmseg - INFO - per class results:
2023-10-26 09:47:06,377 - mmseg - INFO -
+------------+-------+-------+
| Class | IoU | Acc |
+------------+-------+-------+
| background | 31.21 | 100.0 |
| forest | 0.0 | 0.0 |
| shrubs | 0.0 | 0.0 |
+------------+-------+-------+
2023-10-26 09:47:06,377 - mmseg - INFO - Summary:
2023-10-26 09:47:06,377 - mmseg - INFO -
+-------+------+-------+
| aAcc | mIoU | mAcc |
+-------+------+-------+
| 31.21 | 10.4 | 33.33 |
+-------+------+-------+
2023-10-26 09:47:06,380 - mmseg - INFO - Exp name: gta2cs_uda_warm_fdthings_rcs_croppl_a999_daformer_mitb5_s0
2023-10-26 09:47:06,380 - mmseg - INFO - Iter [440/40000] lr: 5.400e-05, eta: 9:40:53, time: 0.986, data_time: 0.013, memory: 9743, aAcc: 0.3121, mIoU: 0.1040, mAcc: 0.3333, IoU.background: 0.3121, IoU.forest: 0.0000, IoU.shrubs: 0.0000, Acc.background: 1.0000, Acc.forest: 0.0000, Acc.shrubs: 0.0000, decode.loss_seg: 0.0001, decode.acc_seg: 63.4068, src.loss_imnet_feat_dist: 0.0487, mix.decode.loss_seg: 0.0001, mix.decode.acc_seg: 79.8197