[WIP] masked_latent_inputs for inpainting pipeline (huggingface#4819)

* add

---------

Co-authored-by: yiyixuxu <yixu310@gmail,com>

src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py

@@ -872,7 +872,11 @@ def prepare_latents(
 
         if return_image_latents or (latents is None and not is_strength_max):
             image = image.to(device=device, dtype=dtype)
-            image_latents = self._encode_vae_image(image=image, generator=generator)
+
+            if image.shape[1] == 4:
+                image_latents = image
+            else:
+                image_latents = self._encode_vae_image(image=image, generator=generator)
 
         if latents is None:
             noise = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
@@ -907,7 +911,11 @@ def prepare_mask_latents(
         mask = mask.to(device=device, dtype=dtype)
 
         masked_image = masked_image.to(device=device, dtype=dtype)
-        masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
+
+        if masked_image.shape[1] == 4:
+            masked_image_latents = masked_image
+        else:
+            masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
 
         # duplicate mask and masked_image_latents for each generation per prompt, using mps friendly method
         if mask.shape[0] < batch_size:

src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py

@@ -293,7 +293,11 @@ def prepare_mask_latents(
         mask = mask.to(device=device, dtype=dtype)
 
         masked_image = masked_image.to(device=device, dtype=dtype)
-        masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
+
+        if masked_image.shape[1] == 4:
+            masked_image_latents = masked_image
+        else:
+            masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
 
         # duplicate mask and masked_image_latents for each generation per prompt, using mps friendly method
         if mask.shape[0] < batch_size:

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -554,7 +554,7 @@ def check_inputs(
         if strength < 0 or strength > 1:
             raise ValueError(f"The value of strength should in [0.0, 1.0] but is {strength}")
 
-        if height % 8 != 0 or width % 8 != 0:
+        if height % self.vae_scale_factor != 0 or width % self.vae_scale_factor != 0:
             raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")
 
         if (callback_steps is None) or (
@@ -622,7 +622,11 @@ def prepare_latents(
 
         if return_image_latents or (latents is None and not is_strength_max):
             image = image.to(device=device, dtype=dtype)
-            image_latents = self._encode_vae_image(image=image, generator=generator)
+
+            if image.shape[1] == 4:
+                image_latents = image
+            else:
+                image_latents = self._encode_vae_image(image=image, generator=generator)
 
         if latents is None:
             noise = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
@@ -670,7 +674,11 @@ def prepare_mask_latents(
         mask = mask.to(device=device, dtype=dtype)
 
         masked_image = masked_image.to(device=device, dtype=dtype)
-        masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
+
+        if masked_image.shape[1] == 4:
+            masked_image_latents = masked_image
+        else:
+            masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
 
         # duplicate mask and masked_image_latents for each generation per prompt, using mps friendly method
         if mask.shape[0] < batch_size:
@@ -715,6 +723,7 @@ def __call__(
         prompt: Union[str, List[str]] = None,
         image: PipelineImageInput = None,
         mask_image: PipelineImageInput = None,
+        masked_image_latents: torch.FloatTensor = None,
         height: Optional[int] = None,
         width: Optional[int] = None,
         strength: float = 1.0,
@@ -914,12 +923,6 @@ def __call__(
         init_image = self.image_processor.preprocess(image, height=height, width=width)
         init_image = init_image.to(dtype=torch.float32)
 
-        mask = self.mask_processor.preprocess(mask_image, height=height, width=width)
-
-        masked_image = init_image * (mask < 0.5)
-
-        mask_condition = mask.clone()
-
         # 6. Prepare latent variables
         num_channels_latents = self.vae.config.latent_channels
         num_channels_unet = self.unet.config.in_channels
@@ -947,8 +950,15 @@ def __call__(
             latents, noise = latents_outputs
 
         # 7. Prepare mask latent variables
+        mask_condition = self.mask_processor.preprocess(mask_image, height=height, width=width)
+
+        if masked_image_latents is None:
+            masked_image = init_image * (mask_condition < 0.5)
+        else:
+            masked_image = masked_image_latents
+
         mask, masked_image_latents = self.prepare_mask_latents(
-            mask,
+            mask_condition,
             masked_image,
             batch_size * num_images_per_prompt,
             height,

src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py

@@ -762,10 +762,16 @@ def prepare_mask_latents(
 
         mask = torch.cat([mask] * 2) if do_classifier_free_guidance else mask
 
-        masked_image_latents = None
+        if masked_image is not None and masked_image.shape[1] == 4:
+            masked_image_latents = masked_image
+        else:
+            masked_image_latents = None
+
         if masked_image is not None:
-            masked_image = masked_image.to(device=device, dtype=dtype)
-            masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
+            if masked_image_latents is None:
+                masked_image = masked_image.to(device=device, dtype=dtype)
+                masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
+
             if masked_image_latents.shape[0] < batch_size:
                 if not batch_size % masked_image_latents.shape[0] == 0:
                     raise ValueError(
@@ -890,6 +896,7 @@ def __call__(
         prompt_2: Optional[Union[str, List[str]]] = None,
         image: PipelineImageInput = None,
         mask_image: PipelineImageInput = None,
+        masked_image_latents: torch.FloatTensor = None,
         height: Optional[int] = None,
         width: Optional[int] = None,
         strength: float = 0.9999,
@@ -1152,7 +1159,9 @@ def denoising_value_valid(dnv):
 
         mask = self.mask_processor.preprocess(mask_image, height=height, width=width)
 
-        if init_image.shape[1] == 4:
+        if masked_image_latents is not None:
+            masked_image = masked_image_latents
+        elif init_image.shape[1] == 4:
             # if images are in latent space, we can't mask it
             masked_image = None
         else:

tests/pipelines/stable_diffusion/test_stable_diffusion_inpaint.py

@@ -252,6 +252,43 @@ def test_stable_diffusion_inpaint_strength_zero_test(self):
         with self.assertRaises(ValueError):
             sd_pipe(**inputs).images
 
+    def test_stable_diffusion_inpaint_mask_latents(self):
+        device = "cpu"
+        components = self.get_dummy_components()
+        sd_pipe = self.pipeline_class(**components).to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        # normal mask + normal image
+        ##  `image`: pil, `mask_image``: pil, `masked_image_latents``: None
+        inputs = self.get_dummy_inputs(device)
+        inputs["strength"] = 0.9
+        out_0 = sd_pipe(**inputs).images
+
+        # image latents + mask latents
+        inputs = self.get_dummy_inputs(device)
+        image = sd_pipe.image_processor.preprocess(inputs["image"]).to(sd_pipe.device)
+        mask = sd_pipe.mask_processor.preprocess(inputs["mask_image"]).to(sd_pipe.device)
+        masked_image = image * (mask < 0.5)
+
+        generator = torch.Generator(device=device).manual_seed(0)
+        image_latents = (
+            sd_pipe.vae.encode(image).latent_dist.sample(generator=generator) * sd_pipe.vae.config.scaling_factor
+        )
+        torch.randn((1, 4, 32, 32), generator=generator)
+        mask_latents = (
+            sd_pipe.vae.encode(masked_image).latent_dist.sample(generator=generator)
+            * sd_pipe.vae.config.scaling_factor
+        )
+        inputs["image"] = image_latents
+        inputs["masked_image_latents"] = mask_latents
+        inputs["mask_image"] = mask
+        inputs["strength"] = 0.9
+        generator = torch.Generator(device=device).manual_seed(0)
+        torch.randn((1, 4, 32, 32), generator=generator)
+        inputs["generator"] = generator
+        out_1 = sd_pipe(**inputs).images
+        assert np.abs(out_0 - out_1).max() < 1e-2
+
 
 class StableDiffusionSimpleInpaintPipelineFastTests(StableDiffusionInpaintPipelineFastTests):
     pipeline_class = StableDiffusionInpaintPipeline

tests/pipelines/stable_diffusion_xl/test_stable_diffusion_xl_inpaint.py

@@ -499,3 +499,35 @@ def test_stable_diffusion_xl_img2img_negative_conditions(self):
             np.abs(image_slice_with_no_neg_conditions.flatten() - image_slice_with_neg_conditions.flatten()).max()
             > 1e-4
         )
+
+    def test_stable_diffusion_xl_inpaint_mask_latents(self):
+        device = "cpu"
+        components = self.get_dummy_components()
+        sd_pipe = self.pipeline_class(**components).to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        # normal mask + normal image
+        ##  `image`: pil, `mask_image``: pil, `masked_image_latents``: None
+        inputs = self.get_dummy_inputs(device)
+        inputs["strength"] = 0.9
+        out_0 = sd_pipe(**inputs).images
+
+        # image latents + mask latents
+        inputs = self.get_dummy_inputs(device)
+        image = sd_pipe.image_processor.preprocess(inputs["image"]).to(sd_pipe.device)
+        mask = sd_pipe.mask_processor.preprocess(inputs["mask_image"]).to(sd_pipe.device)
+        masked_image = image * (mask < 0.5)
+
+        generator = torch.Generator(device=device).manual_seed(0)
+        image_latents = sd_pipe._encode_vae_image(image, generator=generator)
+        torch.randn((1, 4, 32, 32), generator=generator)
+        mask_latents = sd_pipe._encode_vae_image(masked_image, generator=generator)
+        inputs["image"] = image_latents
+        inputs["masked_image_latents"] = mask_latents
+        inputs["mask_image"] = mask
+        inputs["strength"] = 0.9
+        generator = torch.Generator(device=device).manual_seed(0)
+        torch.randn((1, 4, 32, 32), generator=generator)
+        inputs["generator"] = generator
+        out_1 = sd_pipe(**inputs).images
+        assert np.abs(out_0 - out_1).max() < 1e-2