train on one gpu, windows machine about real-esrgan

[xinntao]

For one gpu, there is no need to use distributed training, try

python realesrgan/train.py -opt options/train_realesrnet_x4plus.yml directly:-)

[mohammadmardani]

I tried that too.
it will raise:
raise KeyError(key) from None
KeyError: 'RANK

if I set the os.environ['RANK'] = '0'
then other keyerror's will raise

[xinntao]

Could you please paste more detailed info? from which line raise this error?

[mohammadmardani]

before setting the RANK to 0:

after setting it to 0:

[xinntao]

Please use python realesrgan/train.py -opt options/train_realesrnet_x4plus.yml

do not add --laucher pytorch