Skip to content

Allow dtype to be specified in Flax pipeline #600

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.

Sign up for GitHub

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

Merged
merged 3 commits into from
Sep 21, 2022
Merged

Allow dtype to be specified in Flax pipeline #600

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2d4cfd4
Fix typo in docstring.
pcuenca Sep 21, 2022
79d6047
Allow dtype to be overridden on model load.
pcuenca Sep 20, 2022
dabfe31
Create latents in float32
pcuenca Sep 21, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion src/diffusers/configuration_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -154,9 +154,12 @@ def from_config(cls, pretrained_model_name_or_path: Union[str, os.PathLike], ret

"""
config_dict = cls.get_config_dict(pretrained_model_name_or_path=pretrained_model_name_or_path, **kwargs)

init_dict, unused_kwargs = cls.extract_init_dict(config_dict, **kwargs)

# Allow dtype to be specified on initialization
if "dtype" in unused_kwargs:
init_dict["dtype"] = unused_kwargs.pop("dtype")

model = cls(**init_dict)

if return_unused_kwargs:
Expand Down
4 changes: 2 additions & 2 deletions src/diffusers/pipelines/stable_diffusion/pipeline_flax_stable_diffusion.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
Tokenizer of class
[CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
unet ([`FlaxUNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
scheduler ([`FlaxSchedulerMixin`]):
scheduler ([`SchedulerMixin`]):
A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
[`FlaxDDIMScheduler`], [`FlaxLMSDiscreteScheduler`], or [`FlaxPNDMScheduler`].
safety_checker ([`FlaxStableDiffusionSafetyChecker`]):
Expand Down Expand Up @@ -157,7 +157,7 @@ def __call__(
self.unet.sample_size,
)
if latents is None:
latents = jax.random.normal(prng_seed, shape=latents_shape, dtype=self.dtype)
latents = jax.random.normal(prng_seed, shape=latents_shape, dtype=jnp.float32)
Copy link
Member Author

@pcuenca pcuenca Sep 21, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The alternative to this, I think, would be to prepare the scheduler parameters using the same dtype as the model. We can do that in a follow-up PR.

Copy link
Contributor

@patil-suraj patil-suraj Sep 21, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sounds good, and agree we should check model dtype here.

else:
if latents.shape != latents_shape:
raise ValueError(f"Unexpected latents shape, got {latents.shape}, expected {latents_shape}")
Expand Down