Open Melyns opened 1 month ago
Do you have the T5 clip encoder added to the top bar in "VAE / Text Encoder"? That error is indicating you do not or what you have in there isn't compatible. Double check that.
i dont because the model is all in one. its stand alone, as the name of the file "InclClip" which stands for including clip. https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/sd3_medium_incl_clips.safetensors
You need to use one of the other checkpoints listed there, one that include t5 in the file name. The one you mentioned above, sd3_medium_incl_clips.safetensors
they explicitly says:
_sd3_medium_inclclips.safetensors includes all necessary weights except for the T5XXL text encoder.
Im using 'stableDiffusion3SD3_sd3MediumInclClips.safetensors', it runs fine in comfy, but when i try it in forge I get "AssertionError: You do not have T5 state dict!" any fix to this please?