Linaqruf / kohya-trainer

Adapted from https://note.com/kohya_ss/n/nbf7ce8d80f29 for easier cloning
Apache License 2.0
1.82k stars 296 forks source link

Multiple Errors when initiating SDXL training with most recent update #310

Open mrmeseeks23 opened 9 months ago

mrmeseeks23 commented 9 months ago

Runpod.io VM OS: Linux Ubuntu 22.04 PyTorch: 2.0.1 CUDA: 11.8

Since last git pull, and new requirements documents there seems to be some warnings appearing that I feel may be related to the Tensorflow it installed (just after running git pull then './gui.sh --share --headless'). I have attached images with the errors/warnings. It seems to run okay and produce good results, however I wanted to make sure I wasn't missing something here. Screen Shot 2023-10-14 at 5 26 51 PM Screen Shot 2023-10-14 at 5 26 56 PM

The last warning is almost always there so I am not worried about that one. Register errors befor (ie cuDNN, cuFFT, cuBLAS)