Skip to content

refactor: pipelines readability improvements #622

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 2 commits into from
Sep 23, 2022
Merged

refactor: pipelines readability improvements #622

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8b600cd
refactor: pipelines readability improvements
ryanrussell Sep 23, 2022
f3d6a87
docs: remove todo comment from flax pipeline
ryanrussell Sep 23, 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
8 changes: 4 additions & 4 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 @@ -34,7 +34,7 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
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`]):
Classification module that estimates whether generated images could be considered offsensive or harmful.
Classification module that estimates whether generated images could be considered offensive or harmful.
Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
feature_extractor ([`CLIPFeatureExtractor`]):
Model that extracts features from generated images to be used as inputs for the `safety_checker`.
Expand Down Expand Up @@ -149,7 +149,7 @@ def __call__(
uncond_embeddings = self.text_encoder(uncond_input.input_ids, params=params["text_encoder"])[0]
context = jnp.concatenate([uncond_embeddings, text_embeddings])

# TODO: check it because the shape is different from Pytorhc StableDiffusionPipeline
# TODO: check it because the shape is different from Pytorch StableDiffusionPipeline
Copy link
Member

Choose a reason for hiding this comment

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

Actually I think we can remove this comment now. @patil-suraj

latents_shape = (
batch_size,
self.unet.in_channels,
Expand Down Expand Up @@ -206,9 +206,9 @@ def loop_body(step, args):
# image = jnp.asarray(image).transpose(0, 2, 3, 1)
# run safety checker
# TODO: check when flax safety checker gets merged into main
# safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="np")
# safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="np")
# image, has_nsfw_concept = self.safety_checker(
# images=image, clip_input=safety_cheker_input.pixel_values, params=params["safety_params"]
# images=image, clip_input=safety_checker_input.pixel_values, params=params["safety_params"]
# )
has_nsfw_concept = False

Expand Down
6 changes: 3 additions & 3 deletions src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ class StableDiffusionPipeline(DiffusionPipeline):
A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
[`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
safety_checker ([`StableDiffusionSafetyChecker`]):
Classification module that estimates whether generated images could be considered offsensive or harmful.
Classification module that estimates whether generated images could be considered offensive or harmful.
Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
feature_extractor ([`CLIPFeatureExtractor`]):
Model that extracts features from generated images to be used as inputs for the `safety_checker`.
Expand Down Expand Up @@ -278,8 +278,8 @@ def __call__(
image = image.cpu().permute(0, 2, 3, 1).numpy()

# run safety checker
safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_cheker_input.pixel_values)
safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_checker_input.pixel_values)

if output_type == "pil":
image = self.numpy_to_pil(image)
Expand Down
6 changes: 3 additions & 3 deletions src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,7 +48,7 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
[`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
safety_checker ([`StableDiffusionSafetyChecker`]):
Classification module that estimates whether generated images could be considered offsensive or harmful.
Classification module that estimates whether generated images could be considered offensive or harmful.
Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
feature_extractor ([`CLIPFeatureExtractor`]):
Model that extracts features from generated images to be used as inputs for the `safety_checker`.
Expand Down Expand Up @@ -288,8 +288,8 @@ def __call__(
image = image.cpu().permute(0, 2, 3, 1).numpy()

# run safety checker
safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_cheker_input.pixel_values)
safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_checker_input.pixel_values)

if output_type == "pil":
image = self.numpy_to_pil(image)
Expand Down
6 changes: 3 additions & 3 deletions src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,7 +66,7 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
[`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
safety_checker ([`StableDiffusionSafetyChecker`]):
Classification module that estimates whether generated images could be considered offsensive or harmful.
Classification module that estimates whether generated images could be considered offensive or harmful.
Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
feature_extractor ([`CLIPFeatureExtractor`]):
Model that extracts features from generated images to be used as inputs for the `safety_checker`.
Expand Down Expand Up @@ -328,8 +328,8 @@ def __call__(
image = image.cpu().permute(0, 2, 3, 1).numpy()

# run safety checker
safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_cheker_input.pixel_values)
safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_checker_input.pixel_values)

if output_type == "pil":
image = self.numpy_to_pil(image)
Expand Down
24 changes: 12 additions & 12 deletions src/diffusers/pipelines/stable_diffusion/safety_checker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,20 +48,20 @@ def forward(self, clip_input, images):
# at the cost of increasing the possibility of filtering benign images
adjustment = 0.0

for concet_idx in range(len(special_cos_dist[0])):
concept_cos = special_cos_dist[i][concet_idx]
concept_threshold = self.special_care_embeds_weights[concet_idx].item()
result_img["special_scores"][concet_idx] = round(concept_cos - concept_threshold + adjustment, 3)
if result_img["special_scores"][concet_idx] > 0:
result_img["special_care"].append({concet_idx, result_img["special_scores"][concet_idx]})
for concept_idx in range(len(special_cos_dist[0])):
concept_cos = special_cos_dist[i][concept_idx]
concept_threshold = self.special_care_embeds_weights[concept_idx].item()
result_img["special_scores"][concept_idx] = round(concept_cos - concept_threshold + adjustment, 3)
if result_img["special_scores"][concept_idx] > 0:
result_img["special_care"].append({concept_idx, result_img["special_scores"][concept_idx]})
adjustment = 0.01

for concet_idx in range(len(cos_dist[0])):
concept_cos = cos_dist[i][concet_idx]
concept_threshold = self.concept_embeds_weights[concet_idx].item()
result_img["concept_scores"][concet_idx] = round(concept_cos - concept_threshold + adjustment, 3)
if result_img["concept_scores"][concet_idx] > 0:
result_img["bad_concepts"].append(concet_idx)
for concept_idx in range(len(cos_dist[0])):
concept_cos = cos_dist[i][concept_idx]
concept_threshold = self.concept_embeds_weights[concept_idx].item()
result_img["concept_scores"][concept_idx] = round(concept_cos - concept_threshold + adjustment, 3)
if result_img["concept_scores"][concept_idx] > 0:
result_img["bad_concepts"].append(concept_idx)

result.append(result_img)

Expand Down