loss is not decrease

[GuoQuanhao]

I use 4x1080ti

python -m torch.distributed.launch --nproc_per_node=4 --use_env main.py --model xcit_nano_12_p8 --batch-size 64 --drop-path 0.05 --output_dir ./experiments/xcit_nano_12_p8/ --epochs 100

I got this

Epoch: [1]  [  0/390]  eta: 0:27:35  lr: 0.000001  loss: 6.8866 (6.8866)  time: 4.2456  data: 2.6132  max mem: 6128
Epoch: [1]  [ 10/390]  eta: 0:07:40  lr: 0.000001  loss: 6.9260 (6.9281)  time: 1.2121  data: 0.2377  max mem: 6128
Epoch: [1]  [ 20/390]  eta: 0:06:37  lr: 0.000001  loss: 6.9244 (6.9252)  time: 0.9165  data: 0.0001  max mem: 6128
Epoch: [1]  [ 30/390]  eta: 0:06:09  lr: 0.000001  loss: 6.9226 (6.9256)  time: 0.9227  data: 0.0001  max mem: 6128
Epoch: [1]  [ 40/390]  eta: 0:05:49  lr: 0.000001  loss: 6.9321 (6.9277)  time: 0.9192  data: 0.0001  max mem: 6128
Epoch: [1]  [ 50/390]  eta: 0:05:35  lr: 0.000001  loss: 6.9378 (6.9297)  time: 0.9253  data: 0.0001  max mem: 6128
Epoch: [1]  [ 60/390]  eta: 0:05:21  lr: 0.000001  loss: 6.9406 (6.9320)  time: 0.9267  data: 0.0001  max mem: 6128
Epoch: [1]  [ 70/390]  eta: 0:05:10  lr: 0.000001  loss: 6.9406 (6.9321)  time: 0.9265  data: 0.0001  max mem: 6128
Epoch: [1]  [ 80/390]  eta: 0:04:58  lr: 0.000001  loss: 6.9337 (6.9325)  time: 0.9315  data: 0.0001  max mem: 6128
Epoch: [1]  [ 90/390]  eta: 0:04:48  lr: 0.000001  loss: 6.9340 (6.9328)  time: 0.9339  data: 0.0001  max mem: 6128
Epoch: [1]  [100/390]  eta: 0:04:38  lr: 0.000001  loss: 6.9246 (6.9323)  time: 0.9438  data: 0.0001  max mem: 6128
Epoch: [1]  [110/390]  eta: 0:04:27  lr: 0.000001  loss: 6.9255 (6.9317)  time: 0.9371  data: 0.0001  max mem: 6128
Epoch: [1]  [120/390]  eta: 0:04:17  lr: 0.000001  loss: 6.9293 (6.9317)  time: 0.9224  data: 0.0001  max mem: 6128
Epoch: [1]  [130/390]  eta: 0:04:07  lr: 0.000001  loss: 6.9322 (6.9319)  time: 0.9176  data: 0.0001  max mem: 6128
Epoch: [1]  [140/390]  eta: 0:03:57  lr: 0.000001  loss: 6.9306 (6.9320)  time: 0.9286  data: 0.0001  max mem: 6128
Epoch: [1]  [150/390]  eta: 0:03:47  lr: 0.000001  loss: 6.9294 (6.9319)  time: 0.9332  data: 0.0001  max mem: 6128
Epoch: [1]  [160/390]  eta: 0:03:38  lr: 0.000001  loss: 6.9265 (6.9313)  time: 0.9317  data: 0.0001  max mem: 6128
Epoch: [1]  [170/390]  eta: 0:03:28  lr: 0.000001  loss: 6.9265 (6.9318)  time: 0.9253  data: 0.0001  max mem: 6128
Epoch: [1]  [180/390]  eta: 0:03:18  lr: 0.000001  loss: 6.9412 (6.9319)  time: 0.9212  data: 0.0001  max mem: 6128
Epoch: [1]  [190/390]  eta: 0:03:08  lr: 0.000001  loss: 6.9273 (6.9319)  time: 0.9292  data: 0.0001  max mem: 6128
Epoch: [1]  [200/390]  eta: 0:02:59  lr: 0.000001  loss: 6.9255 (6.9318)  time: 0.9226  data: 0.0001  max mem: 6128
Epoch: [1]  [210/390]  eta: 0:02:49  lr: 0.000001  loss: 6.9305 (6.9317)  time: 0.9275  data: 0.0001  max mem: 6128
Epoch: [1]  [220/390]  eta: 0:02:40  lr: 0.000001  loss: 6.9295 (6.9314)  time: 0.9360  data: 0.0001  max mem: 6128
Epoch: [1]  [230/390]  eta: 0:02:30  lr: 0.000001  loss: 6.9290 (6.9312)  time: 0.9446  data: 0.0001  max mem: 6128
Epoch: [1]  [240/390]  eta: 0:02:21  lr: 0.000001  loss: 6.9229 (6.9305)  time: 0.9421  data: 0.0001  max mem: 6128
Epoch: [1]  [250/390]  eta: 0:02:11  lr: 0.000001  loss: 6.9263 (6.9310)  time: 0.9283  data: 0.0001  max mem: 6128
Epoch: [1]  [260/390]  eta: 0:02:02  lr: 0.000001  loss: 6.9225 (6.9305)  time: 0.9232  data: 0.0001  max mem: 6128
Epoch: [1]  [270/390]  eta: 0:01:52  lr: 0.000001  loss: 6.9225 (6.9307)  time: 0.9220  data: 0.0001  max mem: 6128
Epoch: [1]  [280/390]  eta: 0:01:43  lr: 0.000001  loss: 6.9359 (6.9309)  time: 0.9205  data: 0.0001  max mem: 6128
Epoch: [1]  [290/390]  eta: 0:01:33  lr: 0.000001  loss: 6.9323 (6.9307)  time: 0.9232  data: 0.0001  max mem: 6128
Epoch: [1]  [300/390]  eta: 0:01:24  lr: 0.000001  loss: 6.9245 (6.9304)  time: 0.9327  data: 0.0001  max mem: 6128
Epoch: [1]  [310/390]  eta: 0:01:15  lr: 0.000001  loss: 6.9237 (6.9304)  time: 0.9280  data: 0.0001  max mem: 6128
Epoch: [1]  [320/390]  eta: 0:01:05  lr: 0.000001  loss: 6.9333 (6.9307)  time: 0.9234  data: 0.0001  max mem: 6128
Epoch: [1]  [330/390]  eta: 0:00:56  lr: 0.000001  loss: 6.9372 (6.9308)  time: 0.9362  data: 0.0001  max mem: 6128
Epoch: [1]  [340/390]  eta: 0:00:46  lr: 0.000001  loss: 6.9314 (6.9306)  time: 0.9338  data: 0.0001  max mem: 6128
Epoch: [1]  [350/390]  eta: 0:00:37  lr: 0.000001  loss: 6.9309 (6.9308)  time: 0.9319  data: 0.0001  max mem: 6128
Epoch: [1]  [360/390]  eta: 0:00:28  lr: 0.000001  loss: 6.9258 (6.9307)  time: 0.9357  data: 0.0001  max mem: 6128
Epoch: [1]  [370/390]  eta: 0:00:18  lr: 0.000001  loss: 6.9239 (6.9303)  time: 0.9330  data: 0.0002  max mem: 6128
Epoch: [1]  [380/390]  eta: 0:00:09  lr: 0.000001  loss: 6.9205 (6.9301)  time: 0.9424  data: 0.0001  max mem: 6128
Epoch: [1]  [389/390]  eta: 0:00:00  lr: 0.000001  loss: 6.9206 (6.9300)  time: 0.9389  data: 0.0001  max mem: 6128

[GuoQuanhao]

worked