sd-forge-layerdiffuse icon indicating copy to clipboard operation
sd-forge-layerdiffuse copied to clipboard

how do i run it by diffusers

Open xiekong opened this issue 1 year ago • 4 comments

give some examples,please

xiekong avatar Mar 04 '24 09:03 xiekong

Sharing this thread.

The base layer diffusion model is a SDXL LoRA trained on the modified latent space. I managed to convert the given checkpoint to LoRA and use it with diffusers by calling pipe.load_lora_weights().

But how do you use the other four layer modules:

  • layer_xl_bg2ble.safetensors
  • layer_xl_bgble2fg.safetensors
  • layer_xl_fg2ble.safetensors
  • layer_xl_fgble2bg.safetensors

Its tightly integrated with sd-forge and using things like extra_concat_cond and I have no idea how to integrate them to diffusers.

xiankgx avatar Mar 05 '24 01:03 xiankgx

i am waiting for the official integration plan

xiekong avatar Mar 05 '24 09:03 xiekong

@xiankgx can you share how you used the base model for inference in diffusers?

nighting0le01 avatar Mar 06 '24 19:03 nighting0le01

Also curious how this could work with diffusers

CustodialMatter avatar Mar 27 '24 01:03 CustodialMatter