Text-to-Image
English

i find no difference using lora and not using lora

#1
by ryg81 - opened

Hi,
I tried it in comfy Ui with default workflow and I found no difference in both generations with and without lora are identical. There is no visual difference. Is there any settings need to be done? is there any workflow available to test?
ComfyUI_temp_gzbjp_00006_.png
Here is the generation, it should have workflow embedded.

same with Automatic1111 ... what are the correct settings please? Or can we just run this LoRA with your code on github?

SPO-Diffusion-Models org

Hello, have you tried the checkpoint from here?

image.png
Now it nailed perfectly.
Very huge difference, thanks.

SPO-Diffusion-Models org

You are welcome!

rockeycoss changed discussion status to closed

Hey bro, have one question. Is that anything to do with samplers, as with custom sampler I am not getting any benefit.
please suggest and then close

SPO-Diffusion-Models org

We fine-tuned the model with DDIM sampler and generate the teaser with EulerDiscreteScheduler.

rockeycoss changed discussion status to open

I'm having the same problem with the LoRA. The LoRA would be very useful so we can use it on top of our own models.

Sign up or log in to comment