Impl denoising_start/end for SDXL ControlNet pipeline

src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py

@@ -782,13 +782,40 @@ def prepare_control_image(
         return image
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.StableDiffusionImg2ImgPipeline.get_timesteps
-    def get_timesteps(self, num_inference_steps, strength, device):
+    def get_timesteps(self, num_inference_steps, strength, device, denoising_start=None):
         # get the original timestep using init_timestep
-        init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
+        if denoising_start is None:
+            init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
+            t_start = max(num_inference_steps - init_timestep, 0)
+        else:
+            t_start = 0
 
-        t_start = max(num_inference_steps - init_timestep, 0)
         timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
 
+        # Strength is irrelevant if we directly request a timestep to start at;
+        # that is, strength is determined by the denoising_start instead.
+        if denoising_start is not None:
+            discrete_timestep_cutoff = int(
+                round(
+                    self.scheduler.config.num_train_timesteps
+                    - (denoising_start * self.scheduler.config.num_train_timesteps)
+                )
+            )
+
+            num_inference_steps = (timesteps < discrete_timestep_cutoff).sum().item()
+            if self.scheduler.order == 2 and num_inference_steps % 2 == 0:
+                # if the scheduler is a 2nd order scheduler we might have to do +1
+                # because `num_inference_steps` might be even given that every timestep
+                # (except the highest one) is duplicated. If `num_inference_steps` is even it would
+                # mean that we cut the timesteps in the middle of the denoising step
+                # (between 1st and 2nd devirative) which leads to incorrect results. By adding 1
+                # we ensure that the denoising process always ends after the 2nd derivate step of the scheduler
+                num_inference_steps = num_inference_steps + 1
+
+            # because t_n+1 >= t_n, we slice the timesteps starting from the end
+            timesteps = timesteps[-num_inference_steps:]
+            return timesteps, num_inference_steps
+
         return timesteps, num_inference_steps - t_start
 
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_img2img.StableDiffusionXLImg2ImgPipeline.prepare_latents
@@ -979,6 +1006,14 @@ def do_classifier_free_guidance(self):
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
+    @property
+    def denoising_end(self):
+        return self._denoising_end
+
+    @property
+    def denoising_start(self):
+        return self._denoising_start
+
     @property
     def num_timesteps(self):
         return self._num_timesteps
@@ -995,6 +1030,8 @@ def __call__(
         width: Optional[int] = None,
         strength: float = 0.8,
         num_inference_steps: int = 50,
+        denoising_start: Optional[float] = None,
+        denoising_end: Optional[float] = None,
         guidance_scale: float = 5.0,
         negative_prompt: Optional[Union[str, List[str]]] = None,
         negative_prompt_2: Optional[Union[str, List[str]]] = None,
@@ -1236,6 +1273,8 @@ def __call__(
         self._guidance_scale = guidance_scale
         self._clip_skip = clip_skip
         self._cross_attention_kwargs = cross_attention_kwargs
+        self._denoising_end = denoising_end
+        self._denoising_start = denoising_start
 
         # 2. Define call parameters
         if prompt is not None and isinstance(prompt, str):
@@ -1322,11 +1361,20 @@ def __call__(
             assert False
 
         # 5. Prepare timesteps
+        def denoising_value_valid(dnv):
+            return isinstance(self.denoising_end, float) and 0 < dnv < 1
+
         self.scheduler.set_timesteps(num_inference_steps, device=device)
-        timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, strength, device)
+        timesteps, num_inference_steps = self.get_timesteps(
+            num_inference_steps,
+            strength,
+            device,
+            denoising_start=self.denoising_start if denoising_value_valid else None,
+        )
         latent_timestep = timesteps[:1].repeat(batch_size * num_images_per_prompt)
         self._num_timesteps = len(timesteps)
 
+        add_noise = True if self.denoising_start is None else False
         # 6. Prepare latent variables
         latents = self.prepare_latents(
             image,
@@ -1336,7 +1384,7 @@ def __call__(
             prompt_embeds.dtype,
             device,
             generator,
-            True,
+            add_noise,
         )
 
         # 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
@@ -1395,6 +1443,28 @@ def __call__(
 
         # 8. Denoising loop
         num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
+
+        if (
+            self.denoising_end is not None
+            and self.denoising_start is not None
+            and denoising_value_valid(self.denoising_end)
+            and denoising_value_valid(self.denoising_start)
+            and self.denoising_start >= self.denoising_end
+        ):
+            raise ValueError(
+                f"`denoising_start`: {self.denoising_start} cannot be larger than or equal to `denoising_end`: "
+                + f" {self.denoising_end} when using type float."
+            )
+        elif self.denoising_end is not None and denoising_value_valid(self.denoising_end):
+            discrete_timestep_cutoff = int(
+                round(
+                    self.scheduler.config.num_train_timesteps
+                    - (self.denoising_end * self.scheduler.config.num_train_timesteps)
+                )
+            )
+            num_inference_steps = len(list(filter(lambda ts: ts >= discrete_timestep_cutoff, timesteps)))
+            timesteps = timesteps[:num_inference_steps]
+
         with self.progress_bar(total=num_inference_steps) as progress_bar:
             for i, t in enumerate(timesteps):
                 # expand the latents if we are doing classifier free guidance