search

bmaltais / kohya_ss

Apache License 2.0
8.79k stars 1.14k forks source link

Error with accelerate site package - unknown keys #2584

Open crotalidae75 opened 3 weeks ago

crotalidae75 commented 3 weeks ago

I am trying to run Kohya under Linux (which I have done before successfully), but this new recent install of 24.1.4 is giving me trouble when I try to start training. I get this error complaining about Accelerate:

Traceback (most recent call last):
  File "/home/crotalus/kohya_ss/venv/bin/accelerate", line 8, in <module>
    sys.exit(main())
  File "/home/crotalus/kohya_ss/venv/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 47, in main
    args.func(args)
  File "/home/crotalus/kohya_ss/venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 995, in launch_command
    args, defaults, mp_from_config_flag = _validate_launch_command(args)
  File "/home/crotalus/kohya_ss/venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 852, in _validate_launch_command
    defaults = load_config_from_file(args.config_file)
  File "/home/crotalus/kohya_ss/venv/lib/python3.10/site-packages/accelerate/commands/config/config_args.py", line 72, in load_config_from_file
    return config_class.from_yaml_file(yaml_file=config_file)
  File "/home/crotalus/kohya_ss/venv/lib/python3.10/site-packages/accelerate/commands/config/config_args.py", line 148, in from_yaml_file
    raise ValueError(
ValueError: The config file at /home/crotalus/.cache/huggingface/accelerate/default_config.yaml had unknown keys (['enable_cpu_affinity']), please try upgrading your `accelerate` version or fix (and potentially remove) these keys from your config file.

I'm not really sure what to do here as I can't find much reference to this in regards to Kohya specifically. Any ideas on how to resolve this? I'm running a 4080 Super with a Ryzen 5900X, using Kubuntu 24.04. I'm also using pyenv, set to Python 3.10.14.

Admittedly, I need to delve deeper into Python...

b-fission commented 3 weeks ago

It's throwing an error because the default_config.yaml file has an option that's not recognized. That error should go away if you delete that file, or edit the file to remove the enable_cpu_affinity line from it.