-
Notifications
You must be signed in to change notification settings - Fork 5.3k
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
[docs] Kandinsky guide #4555
[docs] Kandinsky guide #4555
Conversation
The documentation is not available anymore as the PR was closed or merged. |
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.
I like the idea of code-switching very much!
Note that we used to have Kandinsky 2.1 and 2.2 on the same page and then later separated Kandinsky 2.2 into its own page. But I think it would work with the code-switching function :)
Updated with the ControlNet pipelines and opened a PR here huggingface/doc-builder#389 to enable the code-switching. Once approved, I'll update all the code examples with the 2.2 usage for all the tasks. |
This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread. Please note that issues that do not follow the contributing guidelines are likely to be ignored. |
Not stale, waiting on huggingface/doc-builder#389 🙂 |
3a467a1
to
db6ead7
Compare
db6ead7
to
54747b8
Compare
Thanks to @mishig25 for the code-switcher, I've updated the docs with Kandinsky 2.2 examples! 😄 https://moon-ci-docs.huggingface.co/docs/diffusers/pr_4555/en/using-diffusers/kandinsky |
|
||
To use the Kandinsky models for any task, you always start by setting up the prior pipeline to encode the prompt and generate the image embeddings. The prior pipeline also generates `negative_image_embeds` that correspond to the negative prompt `""`. For better results, you can pass an actual `negative_prompt` to the prior pipeline, but this'll increase the effective batch size of the prior pipeline by 2x. | ||
|
||
<hfoptions id="kandinsky-text-to-image"> |
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.
for the <hfoptions> id
, using same id when possible on all hfoptions blocks is better as that would result in changing to one of the options, would change all other hfoptions blocks with same id (probably what the user wants)
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.
awesome, thanks for the tip! 🤝
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.
Clean!
* kandinsky 2.1 first draft * add kandinsky 2.2 * fix identical section headers * try hfoptions syntax * add img2img * add inpaint * add interpolate * fix tag * more cleanups * typo * update hfoptions id * align hfoptions tags
* kandinsky 2.1 first draft * add kandinsky 2.2 * fix identical section headers * try hfoptions syntax * add img2img * add inpaint * add interpolate * fix tag * more cleanups * typo * update hfoptions id * align hfoptions tags
Part of the ongoing effort to make the API docs more lightweight (similar to #4428) for some of the pipelines that are more complex and less straightforward to use. This'll also improve #4290.
Since Kandinsky 2.2 usage is so similar to 2.1, do you think it'd make sense to combine the two on a page? I think this can really avoid duplicating content shared by both pipelines. We can copy this code snippet switching thing from the Datasets docs. The only thing we'd have to make super clear is:
prompt
in the decoding pipeline for Kandinsky 2.2