I have trouble in finding the open-source clip model checkpoint that matches the clip used in stable-diffusion-2-1-base.
You mentioned that you used OpenCLIP-ViT/H as the text encoder.
I tried CLIP-ViT-H-14-laion2B-s32B-b79K provided in HuggingFace and open_clip but found the output embedding of text prompt does not matches the output of text encoder used in stable-diffusion-2-1-base.
Could you tell me where is the checkpoints of the OpenCLIP-ViT/H?
The same issue is proposed in huggingface too but has no official answer yet.
Hi, thanks for your great work!
I have trouble in finding the open-source clip model checkpoint that matches the clip used in stable-diffusion-2-1-base. You mentioned that you used OpenCLIP-ViT/H as the text encoder. I tried CLIP-ViT-H-14-laion2B-s32B-b79K provided in HuggingFace and open_clip but found the output embedding of text prompt does not matches the output of text encoder used in stable-diffusion-2-1-base. Could you tell me where is the checkpoints of the OpenCLIP-ViT/H?
The same issue is proposed in huggingface too but has no official answer yet.
Thank you!