You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for the great work!
Recently I see new models like shuttleai/shuttle-3-diffusion · Hugging Face, which is finetuned on flux-schnell.
Is there any way to use mflux on these models?
I tried with replacing the schnell safetensor's file with shuttle 3's in ~/.cache/huggingface/hub, and ran txt2image.py. This gave me the following errors:
ValueError: Received parameters not in model: model.diffusion_model.single_blocks.22.modulation.lin.bias model.diffusion_model.single_blocks.21.linear2.bias model.diffusion_model.single_blocks.5.modulation.lin.bias model.diffusion_model.double_blocks.14.img_attn.qkv.bias……
The text was updated successfully, but these errors were encountered:
Pushed a fix at #1117. It seems that the checkpoint is not quite compatible and I think the diffusers library loads the model from the transformers folder but the fix should be easier to use since you can just download using huggingface-cli and then export FLUX_SCNHELL=path/to/safetensors and use any model you want.
Thanks for the great work!
Recently I see new models like shuttleai/shuttle-3-diffusion · Hugging Face, which is finetuned on flux-schnell.
Is there any way to use mflux on these models?
I tried with replacing the schnell safetensor's file with shuttle 3's in ~/.cache/huggingface/hub, and ran txt2image.py. This gave me the following errors:
The text was updated successfully, but these errors were encountered: