I tried to run the Lora_Trainer_XL locally (still not successfully, need some more fidgetting). I noticed that the root_dir is set to "~/Loras" when COLAB is not true.
However, Python (at least at the version I'm using, which is 3.12) does not expand ~ to current user home. Instead it just uses the path as is. The directory referred to by this variable is <jupyter_running_dir>/~/Loras, which is way off. The script ended up not find any images although there are files at ~/Loras/<project>/dataset directory.
I tried to run the Lora_Trainer_XL locally (still not successfully, need some more fidgetting). I noticed that the
root_dir
is set to"~/Loras"
whenCOLAB
is not true.However, Python (at least at the version I'm using, which is 3.12) does not expand
~
to current user home. Instead it just uses the path as is. The directory referred to by this variable is<jupyter_running_dir>/~/Loras
, which is way off. The script ended up not find any images although there are files at~/Loras/<project>/dataset
directory.A fix would be, to replace:
with
This has been tested on a Linux system (Manjaro Linux). Haven't tried on Windows yet though.