meaningful documentation

optimum/onnxruntime/modeling_diffusion.py

@@ -34,6 +34,7 @@
 from diffusers.utils import CONFIG_NAME
 from huggingface_hub import snapshot_download
 from transformers import CLIPFeatureExtractor, CLIPTokenizer
+from transformers.file_utils import add_end_docstrings
 
 import onnxruntime as ort
 
@@ -51,7 +52,7 @@
     DIFFUSION_MODEL_VAE_DECODER_SUBFOLDER,
     DIFFUSION_MODEL_VAE_ENCODER_SUBFOLDER,
 )
-from .modeling_ort import ORTModel
+from .modeling_ort import ONNX_MODEL_END_DOCSTRING, ORTModel
 from .utils import (
     _ORT_TO_NP_TYPE,
     ONNX_WEIGHTS_NAME,
@@ -531,18 +532,31 @@ def forward(self, sample: np.ndarray):
         return outputs
 
 
+@add_end_docstrings(ONNX_MODEL_END_DOCSTRING)
 class ORTStableDiffusionPipeline(StableDiffusionPipelineMixin, ORTStableDiffusionPipelineBase):
+    """
+    ONNX Runtime-powered stable diffusion pipeline corresponding to [diffusers.StableDiffusionPipeline](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline).
+    """
+
     pass
 
 
-class ORTStableDiffusionImg2ImgPipeline(ORTStableDiffusionPipelineBase, StableDiffusionImg2ImgPipelineMixin):
-    def __call__(self, *args, **kwargs):
-        return StableDiffusionImg2ImgPipelineMixin.__call__(self, *args, **kwargs)
+@add_end_docstrings(ONNX_MODEL_END_DOCSTRING)
+class ORTStableDiffusionImg2ImgPipeline(StableDiffusionImg2ImgPipelineMixin, ORTStableDiffusionPipelineBase):
+    """
+    ONNX Runtime-powered stable diffusion pipeline corresponding to [diffusers.StableDiffusionImg2ImgPipeline](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/img2img#diffusers.StableDiffusionImg2ImgPipeline).
+    """
 
+    pass
 
-class ORTStableDiffusionInpaintPipeline(ORTStableDiffusionPipelineBase, StableDiffusionInpaintPipelineMixin):
-    def __call__(self, *args, **kwargs):
-        return StableDiffusionInpaintPipelineMixin.__call__(self, *args, **kwargs)
+
+@add_end_docstrings(ONNX_MODEL_END_DOCSTRING)
+class ORTStableDiffusionInpaintPipeline(StableDiffusionInpaintPipelineMixin, ORTStableDiffusionPipelineBase):
+    """
+    ONNX Runtime-powered stable diffusion pipeline corresponding to [diffusers.StableDiffusionInpaintPipeline](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/inpaint#diffusers.StableDiffusionInpaintPipeline).
+    """
+
+    pass
 
 
 class ORTStableDiffusionXLPipelineBase(ORTStableDiffusionPipelineBase):
@@ -584,11 +598,19 @@ def __init__(
         self.watermark = StableDiffusionXLWatermarker()
 
 
-class ORTStableDiffusionXLPipeline(ORTStableDiffusionXLPipelineBase, StableDiffusionXLPipelineMixin):
-    def __call__(self, *args, **kwargs):
-        return StableDiffusionXLPipelineMixin.__call__(self, *args, **kwargs)
+@add_end_docstrings(ONNX_MODEL_END_DOCSTRING)
+class ORTStableDiffusionXLPipeline(StableDiffusionXLPipelineMixin, ORTStableDiffusionXLPipelineBase):
+    """
+    ONNX Runtime-powered stable diffusion pipeline corresponding to [diffusers.StableDiffusionXLPipeline](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLPipeline).
+    """
 
+    pass
 
-class ORTStableDiffusionXLImg2ImgPipeline(ORTStableDiffusionXLPipelineBase, StableDiffusionXLImg2ImgPipelineMixin):
-    def __call__(self, *args, **kwargs):
-        return StableDiffusionXLImg2ImgPipelineMixin.__call__(self, *args, **kwargs)
+
+@add_end_docstrings(ONNX_MODEL_END_DOCSTRING)
+class ORTStableDiffusionXLImg2ImgPipeline(StableDiffusionXLImg2ImgPipelineMixin, ORTStableDiffusionXLPipelineBase):
+    """
+    ONNX Runtime-powered stable diffusion pipeline corresponding to [diffusers.StableDiffusionXLImg2ImgPipeline](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLImg2ImgPipeline).
+    """
+
+    pass

optimum/pipelines/diffusers/pipeline_stable_diffusion.py

@@ -135,8 +135,8 @@ def _encode_prompt(
     def check_inputs(
         self,
         prompt: Union[str, List[str]],
-        height: Optional[int],
-        width: Optional[int],
+        height: int,
+        width: int,
         callback_steps: int,
         negative_prompt: Optional[str] = None,
         prompt_embeds: Optional[np.ndarray] = None,
@@ -226,10 +226,10 @@ def __call__(
             prompt (`Optional[Union[str, List[str]]]`, defaults to None):
                 The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
                 instead.
-            height (`Optional[int]`, defaults to None):
-                The height in pixels of the generated image.
-            width (`Optional[int]`, defaults to None):
-                The width in pixels of the generated image.
+            height (`Optional[int]`, defaults to `None`):
+                The height in pixels of the generated image. If `None`, defaults to `self.unet.config.sample_size * self.vae_scale_factor`
+            width (`Optional[int]`, defaults to `None`):
+                The width in pixels of the generated image. If `None`, defaults to `self.unet.config.sample_size * self.vae_scale_factor`
             num_inference_steps (`int`, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
@@ -248,7 +248,7 @@ def __call__(
             eta (`float`, defaults to 0.0):
                 Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
                 [`schedulers.DDIMScheduler`], will be ignored for others.
-            generator (`Optional[np.random.RandomState]`, defaults to `None`)::
+            generator (`Optional[np.random.RandomState]`, defaults to `None`):
                 A np.random.RandomState to make generation deterministic.
             latents (`Optional[np.ndarray]`, defaults to `None`):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image

optimum/pipelines/diffusers/pipeline_stable_diffusion_inpaint.py

@@ -55,8 +55,8 @@ class StableDiffusionInpaintPipelineMixin(StableDiffusionPipelineMixin):
     def check_inputs(
         self,
         prompt: Union[str, List[str]],
-        height: Optional[int],
-        width: Optional[int],
+        height: int,
+        width: int,
         callback_steps: int,
         negative_prompt: Optional[str] = None,
         prompt_embeds: Optional[np.ndarray] = None,
@@ -133,9 +133,9 @@ def __call__(
             mask_image (`PIL.Image.Image`):
                 `Image`, or tensor representing a masked image batch which will be upscaled.
             height (`Optional[int]`, defaults to None):
-                The height in pixels of the generated image.
+                The height in pixels of the generated image. If `None`, defaults to `self.unet.config.sample_size * self.vae_scale_factor`.
             width (`Optional[int]`, defaults to None):
-                The width in pixels of the generated image.
+                The width in pixels of the generated image. If `None`, defaults to `self.unet.config.sample_size * self.vae_scale_factor`.
             num_inference_steps (`int`, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.

optimum/pipelines/diffusers/pipeline_stable_diffusion_xl.py

@@ -293,10 +293,10 @@ def __call__(
             prompt (`Optional[Union[str, List[str]]]`, defaults to None):
                 The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
                 instead.
-            height (`Optional[int]`, defaults to None):
-                The height in pixels of the generated image.
-            width (`Optional[int]`, defaults to None):
-                The width in pixels of the generated image.
+            height (`Optional[int]`, defaults to `None`):
+                The height in pixels of the generated image. If `None`, defaults to `self.unet.config.sample_size * self.vae_scale_factor`.
+            width (`Optional[int]`, defaults to `None`):
+                The width in pixels of the generated image. If `None`, defaults to `self.unet.config.sample_size * self.vae_scale_factor`.
             num_inference_steps (`int`, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
@@ -315,7 +315,7 @@ def __call__(
             eta (`float`, defaults to 0.0):
                 Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
                 [`schedulers.DDIMScheduler`], will be ignored for others.
-            generator (`Optional[np.random.RandomState]`, defaults to `None`)::
+            generator (`Optional[np.random.RandomState]`, defaults to `None`):
                 A np.random.RandomState to make generation deterministic.
             latents (`Optional[np.ndarray]`, defaults to `None`):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image

optimum/pipelines/diffusers/pipeline_stable_diffusion_xl_img2img.py

@@ -325,7 +325,7 @@ def __call__(
             eta (`float`, defaults to 0.0):
                 Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
                 [`schedulers.DDIMScheduler`], will be ignored for others.
-            generator (`Optional[np.random.RandomState]`, defaults to `None`)::
+            generator (`Optional[np.random.RandomState]`, defaults to `None`):
                 A np.random.RandomState to make generation deterministic.
             latents (`Optional[np.ndarray]`, defaults to `None`):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image