openai / CLIP

CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Is the clip preprocessor differentiable ?

aradhyamathur opened this issue · comments

Hi, I had a query pertaining to the clip preprocessor , say i have an image that is generated by some other network, if i process it using clip preprocessor will i be able to backprop the gradients back to the original model ?

I have exactly the same question. It seems it can't. Do you have any idea how to make it differentiable?