-
Notifications
You must be signed in to change notification settings - Fork 6.1k
Support tiled encode/decode for AutoencoderTiny
#4627
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
AutoencoderTiny
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice this looks good! Thank you!
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sweet!
Let's make sure the CI is green before merging :-) |
I think the CI should be green this time. Can you approve the worflows? @sayakpaul |
Thanks for your amazing contribution! |
…port. Now available in diffusers 0.21: huggingface/diffusers#4627
* Impl tae slicing and tiling * add tae tiling test * add parameterized test * formatted code * fix failed test * style docs
* Impl tae slicing and tiling * add tae tiling test * add parameterized test * formatted code * fix failed test * style docs
What does this PR do?
AutoencoderTiny
Though
taesd
reduce VRAM cost compared with vae, it still cost large amount VRAM for encode/decode extremely large image since memory scales linearly with size of the latents.According to my test, tiled-tae decoding only costs 1.2G VRAM with 64x64 tile.
The tile encode/decode are modified from the code presented in madebyollin/taesd#8.
Who can review?
Can you take a look at this? @sayakpaul