SD3 training error
GoofyAi opened this issue · comments
hello i was trying the SD3 branch for testing sd3 training using https://github.com/bmaltais/kohya_ss/tree/sd3 this branch.
it installed right and I got all the sd3 settings right. but when I start the training it would stop the training with an error of "AssertionError: training without text encoder cache is not supported currently / text encoderのキャッシュなしの学習は現 在サポートされていません"
check it out:
but here are my settings on the ui and the toml config. as you can see i have the option enabled on both side but im still getting the error.
can anyone help me with it?
Could you please share the full command line and the .toml?
Could you please share the full command line and the
.toml?
found out that the GUI had issue and he patched it. but still there are some issues. also there is no lora training support for sd3 as he said your script don't support them . i can link you to that discussion bmaltais/kohya_ss#2631 (comment)
also hoping to see you add support for hunyuan dit lora training support as the given process by the hunyuan team I keep getting error for --dynamo-mode
Could you please share the full command line and the
.toml?found out that the GUI had issue and he patched it. but still there are some issues. also there is no lora training support for sd3 as he said your script don't support them . i can link you to that discussion bmaltais/kohya_ss#2631 (comment)
also hoping to see you add support for hunyuan dit lora training support as the given process by the hunyuan team I keep getting error for --dynamo-mode
also hoping for lora training sd3 support
I will add LoRA training when SD3 training is verified. Because the model, noise and loss function etc are same for SD3 and SD3 LoRA.
The same problem
I will add LoRA training when SD3 training is verified. Because the model, noise and loss function etc are same for SD3 and SD3 LoRA.
Hi, any news ?