A note about BrushData!

[zhanjiahui]

I download brushdata from the following link:
Brushdata on huggingface from ramdom123123

Then my training was abnormally terminated because one of the tar files was corrupted.
I asked ChatGPT to write a script to help me detect incorrect tar files:

  import tarfile
  import os
  
  tar_dir = '/your/path/BrushData/'
  
  for tar_file in os.listdir(tar_dir):
      if not tar_file.endswith('.tar'):
          continue
      path = os.path.join(tar_dir, tar_file)
      try:
          with tarfile.open(path, 'r') as tar:
              print(path)
              tar.getmembers()
      except tarfile.ReadError:
          print(f"{tar_file} is corrupted.")

Finally I found 00185.tar was corrupted.

Additionally, according to a discussion, 00182.tar file is missing too.