search

jaketae / storyteller

Multimodal AI Story Teller, built with Stable Diffusion, GPT, and neural text-to-speech
MIT License
482 stars 64 forks source link

Is it possible to successfully execute code in colab? #15

Closed KirillKukharev closed 1 year ago

KirillKukharev commented 1 year ago

I've got an error "The session has been terminated because the amount of RAM is completely exhausted." The same problem with GPU. Are there ways to optimize memory in Story Teller for colab?

jaketae commented 1 year ago

Hello @KirillKukharev, thanks for opening this issue. I've been meaning to create a Colab demo notebook but got carried away by other things.

One thing that immediately comes to mind is using accelerate. IIRC, simply pip install accelerate should make things better due to their default setting. However, your mileage may vary depending on the configuration and generation arguments.

Story Teller is already generating one image at a time instead of batching, so I would say that it is already erring on the side of being slow while saving on memory. I'm curious to see at which stage memory is running out (File I/O, image generation, TTS). Sorry that I don't have a definitive answer; please keep me posted if you have any updates, and I'll do the same. Thanks!

stale[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.