Are you planning to play with diffusers to generate images from text?
If the answer is yes, then continue reading my experience :)
Diffusers library: https://github.com/huggingface/diffusers.
First, we are going to use the branch ldm-bert from the transformers’ fork made by patil-suraj, because it has the model that will be used by the latent diffusion.
Checking the versions of transformers and diffusers …
We are going to use the class DiffusionPipeline to load the latent diffusion model.
Let’s create a function to show the generated images during the inference steps.
And now we can see the process of creating some kitties :)
If you want to play with the Colab, here it is.
Play with it and share your findings 🤗