Unsatisfactory training results on SwinIR

[DX3906ghh]

After I prepared the dataset and environment, I followed the steps of BasicSR to train SwinIR, but I didn’t get good results. On x4, the psnr of set5 only has 30.6348.but the result of paper is 32.72

[clodiamery]

@DX3906ghh , Can I get on some help from you how can train the code on my data, please!

[DX3906ghh]

@DX3906ghh , Can I get on some help from you how can train the code on my data, please!，请问您可以帮助我如何在我的数据上训练代码吗？

of course OK

[clodiamery]

Thank you so much for your reply.

1- Please, what is the computer resources that I need to train it on my own dataset if I have 800 images 250*250 gray?

2- I upload the code on my Drive to run it by using google colab.
I updated the json file to be like this:

"datasets": {
    "train": {
      "name": "train_dataset"           // just name
      , "dataset_type": "jpeg"         // "dncnn" | "dnpatch" | "fdncnn" | "ffdnet" | "sr" | "srmd" | "dpsr" | "plain" | "plainpatch" | "jpeg"
      , "dataroot_H": "train/hr"// path of H training dataset. DIV2K (800 training images) + Flickr2K (2650 images) + BSD500 (400 training&testing images) + WED(4744 images) in SwinIR
      , "dataroot_L": "train/lr"   //null      // path of L training dataset

And I used this command to run it :
!python -m torch.distributed.launch --nproc_per_node=1 --master_port=1234 /content/drive/MyDrive/colab_notebooks/SwinIR-main/main_train_psnr.py --opt /content/drive/MyDrive/colab_notebooks/SwinIR-main/options/swinir/train_swinir_car_jpeg.json --dist True

But I have this error:

/usr/local/lib/python3.10/dist-packages/torch/distributed/launch.py:208: FutureWarning: The module torch.distributed.launch is deprecated
and will be removed in future. Use torchrun.
Note that --use-env is set by default in torchrun.
If your script expects `--local-rank` argument to be set, please
change it to read from `os.environ['LOCAL_RANK']` instead. See 
https://pytorch.org/docs/stable/distributed.html#launch-utility for 
further instructions

  main()
----project SwinIR------
usage: main_train_psnr.py [-h] [--opt OPT] [--launcher LAUNCHER] [--local_rank LOCAL_RANK]
                          [--dist DIST] [--task TASK] [--scale SCALE] [--noise NOISE]
                          [--jpeg JPEG] [--folder_lq FOLDER_LQ] [--folder_gt FOLDER_GT]
                          [--model_save_dir MODEL_SAVE_DIR] [--chart_save_dir CHART_SAVE_DIR]
main_train_psnr.py: error: unrecognized arguments: --local-rank=0
E0114 10:51:36.380000 3268 torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 2) local_rank: 0 (pid: 3280) of binary: /usr/bin/python3
Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/usr/local/lib/python3.10/dist-packages/torch/distributed/launch.py", line 208, in <module>
    main()
  File "/usr/local/lib/python3.10/dist-packages/typing_extensions.py", line 2853, in wrapper
    return arg(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/torch/distributed/launch.py", line 204, in main
    launch(args)
  File "/usr/local/lib/python3.10/dist-packages/torch/distributed/launch.py", line 189, in launch
    run(args)
  File "/usr/local/lib/python3.10/dist-packages/torch/distributed/run.py", line 910, in run
    elastic_launch(
  File "/usr/local/lib/python3.10/dist-packages/torch/distributed/launcher/api.py", line 138, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/usr/local/lib/python3.10/dist-packages/torch/distributed/launcher/api.py", line 269, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
============================================================
/content/drive/MyDrive/colab_notebooks/SwinIR-main/main_train_psnr.py FAILED
------------------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2025-01-14_10:51:36
  host      : 443d37f8eccf
  rank      : 0 (local_rank: 0)
  exitcode  : 2 (pid: 3280)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
============================================================

[mearcla]

@DX3906ghh, Please I have question, I trained train_swinir_car_jpeg.json on my dataset, the model that I got are:

• m1_E.pth
• m2_G.pth
• m3_optimizerG.pth
  What is the difference between them? and which one I should use to enhance my images?

[tingwenabc]

准备好数据集和环境后，我按照 BasicSR 的步骤训练了 SwinIR，但没有得到很好的结果。在 x4 上，set5 的 psnr 只有 30.6348，但 paper 的结果是 32.72

你好 有些问题需要向你请教一下 希望得到你的回复 @DX3906ghh