Update Cascade documentation #7257
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
sayakpaul
reviewed
Mar 8, 2024
sayakpaul
reviewed
Mar 8, 2024
Comment on lines
52
to
53
| prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", variant="bf16", torch_dtype=torch.bfloat16) | ||
| decoder = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade", variant="bf16", torch_dtype=torch.float16) |
There was a problem hiding this comment.
Can we make it clear here why we're loading a pipeline in float16 and another bfloat16? Also, let's maybe mention the minimum torch version requirement in order to run things in bfloat16?
sayakpaul
reviewed
Mar 8, 2024
Comment on lines
104
to
112
| <Tip warning={true}> | ||
|
|
||
| There are some restrictions on data types that can be used with the Stable Cascade models. The official checkpoints for the `StableCascadePriorPipeline` do not support the `torch.float16` data type. Please use `torch.bfloat16` instead. | ||
|
|
||
| In order to use the `torch.bfloat16` data type with the `StableCascadeDecoderPipeline` you need to have PyTorch 2.2.0 or higher installed. This also means that using the `StableCascadeCombinedPipeline` with `torch.bfloat16` requires PyTorch 2.2.0 or higher, since it calls the `StableCascadeDecoderPipeline` internally. | ||
|
|
||
| If it is not possible to install PyTorch 2.2.0 or higher in your environment, the `StableCascadeDecoderPipeline` can be used on its own with the `torch.float16` data type. You can download the full precision or `bf16` variant weights for the pipeline and cast the weights to `torch.float16`. | ||
|
|
||
| </Tip> |
There was a problem hiding this comment.
Do we want to maybe move this section up?
sayakpaul
approved these changes
Mar 8, 2024
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
yiyixuxu
approved these changes
Mar 8, 2024
src/diffusers/pipelines/stable_cascade/pipeline_stable_cascade_combined.py
Show resolved
Hide resolved
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
There are some issues related to dtypes and torch versions with the Stable Cascade model. This PR:
torch.bfloat16dtype can be used with this model.Fixes # (issue)
Before submitting
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.