Create new template from runpod sdxl github release 1.2.0
I see that the github release has been updated here:
https://github.com/runpod-workers/worker-sdxl and we are now at release 1.2.0
But if I try to use on of the ready templates it starts download the old 1.0.4
https://www.runpod.io/console/serverless/quick-deploy
How can i create a template using the last one?
(I see that there are some merge request for that, can you guys create a new release?)
GitHub
GitHub - runpod-workers/worker-sdxl: RunPod worker for Stable Diffu...
RunPod worker for Stable Diffusion XL. Contribute to runpod-workers/worker-sdxl development by creating an account on GitHub.
10 Replies
@danielzanchi you could create a custom template and set the version to 1.2.0 yourself. Because looking at the github-workflow for this image, the
latest-tag is not pushed to docker-hub, meaning that you explicility have to define the version.
So I assume that the template is hard-coded on the 1.0.4 (but I can't confirm, as I have the maximum amount of workers reached and can't deploy the quick-deploy template myself)How could I explicility to define the version to get the last one that I see on github?
I already created a new template from the 1.2.0 succesfully
And even with the 1.2.0 in your custom template it's not working?
Looks like we need to update the quick-deploy version, I'll get that updated shortly. Testing a 1.2.1 release which exposes an extra parameter.
@merrell.io just out of interest: Why are you not using
latest for these templates? Or let the user decide which version they want? Should I add this as "feedback"?Consistency with version control, appending :latest is still a tag.
why there's a limit in output images? o would like to create 4 images.
https://github.com/runpod-workers/worker-sdxl/blob/main/src/rp_schemas.py
line 60
GitHub
worker-sdxl/src/rp_schemas.py at main · runpod-workers/worker-sdxl
RunPod worker for Stable Diffusion XL. Contribute to runpod-workers/worker-sdxl development by creating an account on GitHub.
Probably due the the resolution of the SDXL images. This obviously causes the file size of the image to increase and the size of the response is limited unless you use S3 compatible object storage for storing the output images.
they handle this on replicate 😦
You can fork it and make it do whatever you want, you have the link ti the source