jbilcke-hf/ai-comic-factory · How to upgrade from CPU to GPU for faster generation and iteration?

Nov 10, 2023

I just made a comic. It took a lot of trial and error which took a lot of time. So I tried to find how to upgrade CPU to save time but I could not find any.
Then I duplicated this space to my profile and went to my spaces and upgraded to GPU but it didn't seem to work. It was throwing a lot of errors.
I'm new here so I don't know how everything works. If you could please help me it would be really awesome! Thank you :)

jbilcke-hf

Owner Nov 12, 2023

•

edited Nov 12, 2023

Hi,

you normally do not need to use pay for a GPU on this space since all the generation (for LLM and stable diffusion) is done on external servers!

the best approach is to:

check that you have use a PRO Hugging Face account (it's a modest monthly cost, and this gives you access to unlimited SDXL and LLM for llama-2)
duplicate the AI Comic Factory space
in the popup (to configure the environment variable) put your Hugging Face API token https://huggingface.co/docs/hub/security-tokens
in the same popup, type INFERENCE_API for the "rendering engine", and INFERENCE_API for the "LLM engine"

oneyellowpeacock

Nov 13, 2023

Thankk you so much for the detailed steps!

By external servers do you mean that you are bearing the cost of all the generations?

I will definitely check out Hugging Face Pro. I already have a 3070 laptop but both llama-2 and SDXL seem to be run quite slow in it. So a pro account definitely might help in that regard as well :)
I didn't know about putting the tokens and inference_api. I'll do that and let you know if all worked.

Again Thank you for this awesome tool :)