add T5 lora train and infer for FluxPipeline

[huangjun12]

train and infer code of transformer lora + clip text encoder lora + T5 text encoder lora with FluxPipeline and FluxFillPipeline

[asomoza]

Hi, is this a feature request?

AFAIK FluxFill can be finetuned the same as Flux so no need to separate both.

The training code is an example you can adapt to your needs, and maybe we will add it in the future but it's not a priority right now since most people won't have the hardware required for this, making it useful for only a few.

Also I suggest that if you want to do some more serious finetune with the text encoders, to use a specialized training library, sadly I don't know which one has support for training all the text encoders.