Flux.1 redux error i received.

#4
by yeah25463 - opened

I have been getting this error (CLIPVisionLoader
Error(s) in loading state_dict for CLIPVisionModelProjection:
size mismatch for vision_model.embeddings.patch_embedding.weight: copying a param with shape torch.Size([1152, 3, 14, 14]) from checkpoint, the shape in current model is torch.Size([1024, 3, 14, 14]).) all day, I tried to troubleshoot it for hours, but I couldn't fix it.
I am using flux.1 def fp8, also t5xxl fp8 em3fn, Clip l and vae, I tried it with gguf models and shnell models too, nothing worked and I got the same error again and again. I tried it with the blonde anime girl with massive ears, I got the workflow from flux website, I believe. clip vision went into clip vision and redux model went into model styles folder.
(siglip-so400m-patch14-384.safetensors) 3.4 GB and this file (sigclip_vision_patch14_384.safetensors) 836 MB
a little help please

yep, I'm getting this error too. would be cool to see a solution.

same error here Error(s) in loading state_dict for CLIPVisionModelProjection:
size mismatch for vision_model.embeddings.patch_embedding.weight: copying a param with shape torch.Size([1152, 3, 14, 14])

Manager wasnt updating comfy.
Had to run update_comfyui.bat

getting this error also. Ive tried updating, that didn't work

I have same issue

Same issue as well

I downloaded the Desktop version of Comfy UI and it works!

CLIPVisionLoader
Error(s) in loading state_dict for CLIPVisionModelProjection:
size mismatch for vision_model.embeddings.patch_embedding.weight: copying a param with shape torch.Size([1152, 3, 14, 14]) from checkpoint, the shape in current model is torch.Size([1024, 3, 14, 14]).
size mismatch for vision_model.embeddings.position_embedding.weight: copying a param with shape torch.Size([729, 1152]) from checkpoint, the shape in current model is torch.Size([257, 1024]).

θž’εΉ•ζ“·ε–η•«ι’ 2024-12-08 085609.png

Sign up or log in to comment