[Early Access] SDXL + Refiner in Google Colab

It turns out that the Stable Diffusion developers have closed free access to the ClipDrop service with the SDXl model. I used it quite often when I needed to quickly generate something. 

I have decided to create a Colab notebook that performs this exact task without a user interface. You simply input a prompt and a negative prompt, and Google Colab will generate an image for you without any problems or restrictions. The most important thing is that this does not violate Colab's rules and will work very reliably and for a long time. 

Perhaps it's worth adding some basic settings, such as resolution, steps, samplers, custom models, and so on. Let me know what you think about this.