Simo Ryu

So what you are doing is essentially what Im doing in `lora_add` cli, in mode `upl` (unet + lora). Can you check if that works for you?

```python loaded_pipeline = StableDiffusionPipeline.from_pretrained( path_1, ).to("cpu") weight_apply_lora(loaded_pipeline.unet, torch.load(path_2), alpha=alpha) if output_path.endswith(".pt"): output_path = output_path[:-3] loaded_pipeline.save_pretrained(output_path) ```

Seems pretty much the same logic for me, not sure why yours wouldn't work

By the way, nice work implementing to automatic's webui. I appreciate your work!

So, I think this issue is resolved via #18 ?

Wait what? I didn't know this even existed. You guys have adapted LoRA here?

Lora doesn't train with text encoder yet!

@raymondgp I've found that using large learning rate helps a lot! Use around 5e-5 ~ 4e-4 to see meaningful effect!

@raymondgp Now training with text-encoder is available!

> @cloneofsimo Indeed! I ran the first 10K with Lora and text training, 4e-4, my subject is there, not fully learned, but I will work on improving this. > >...