SD3 unnecessary train_t5xxl check results in error

[dylanprins]

Hi!

In sd3_train_network.py we have the following function:

def post_process_network(self, args, accelerator, network, text_encoders, unet): # check t5xxl is trained or not self.train_t5xxl = network.train_t5xxl

But network has no train_t5xxl and self.train_t5xxl is already initialised with False in the beginning. It now crashes there when i try to train my LoRA. If i'm correct this flag is only for full finetuning?

Should this function simply be removed in sd3_train_network since it is only for a full fine tune?

[dylanprins]

Please update the README for SD3 training, it states you need to use the networks.lora instead of networks.lora_sd3 module. That's why the network wasn't initialised properly.