[Torch 2.0 compile] Fix more torch compile breaks (huggingface#3313)

* Fix more torch compile breaks

* add tests

* Fix all

* fix controlnet

* fix more

* Add Horace He as co-author.
>
>
Co-authored-by: Horace He <[email protected]>

* Add Horace He as co-author.

Co-authored-by: Horace He <[email protected]>

---------

Co-authored-by: Horace He <[email protected]>

Author: 2 people

Diff for: src/diffusers/models/controlnet.py

@@ -498,7 +498,7 @@ def forward(
         # timesteps does not contain any weights and will always return f32 tensors
         # but time_embedding might actually be running in fp16. so we need to cast here.
         # there might be better ways to encapsulate this.
-        t_emb = t_emb.to(dtype=self.dtype)
+        t_emb = t_emb.to(dtype=sample.dtype)
 
         emb = self.time_embedding(t_emb, timestep_cond)
 
@@ -517,7 +517,7 @@ def forward(
 
         controlnet_cond = self.controlnet_cond_embedding(controlnet_cond)
 
-        sample += controlnet_cond
+        sample = sample + controlnet_cond
 
         # 3. down
         down_block_res_samples = (sample,)
@@ -551,21 +551,22 @@ def forward(
 
         for down_block_res_sample, controlnet_block in zip(down_block_res_samples, self.controlnet_down_blocks):
             down_block_res_sample = controlnet_block(down_block_res_sample)
-            controlnet_down_block_res_samples += (down_block_res_sample,)
+            controlnet_down_block_res_samples = controlnet_down_block_res_samples + (down_block_res_sample,)
 
         down_block_res_samples = controlnet_down_block_res_samples
 
         mid_block_res_sample = self.controlnet_mid_block(sample)
 
         # 6. scaling
         if guess_mode:
-            scales = torch.logspace(-1, 0, len(down_block_res_samples) + 1)  # 0.1 to 1.0
-            scales *= conditioning_scale
+            scales = torch.logspace(-1, 0, len(down_block_res_samples) + 1, device=sample.device)  # 0.1 to 1.0
+
+            scales = scales * conditioning_scale
             down_block_res_samples = [sample * scale for sample, scale in zip(down_block_res_samples, scales)]
-            mid_block_res_sample *= scales[-1]  # last one
+            mid_block_res_sample = mid_block_res_sample * scales[-1]  # last one
         else:
             down_block_res_samples = [sample * conditioning_scale for sample in down_block_res_samples]
-            mid_block_res_sample *= conditioning_scale
+            mid_block_res_sample = mid_block_res_sample * conditioning_scale
 
         if self.config.global_pool_conditions:
             down_block_res_samples = [

Diff for: src/diffusers/models/unet_2d_condition.py

@@ -740,7 +740,7 @@ def forward(
                 down_block_res_samples, down_block_additional_residuals
             ):
                 down_block_res_sample = down_block_res_sample + down_block_additional_residual
-                new_down_block_res_samples += (down_block_res_sample,)
+                new_down_block_res_samples = new_down_block_res_samples + (down_block_res_sample,)
 
             down_block_res_samples = new_down_block_res_samples
 

Diff for: src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -457,7 +457,7 @@ def decode_latents(self, latents):
             FutureWarning,
         )
         latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents).sample
+        image = self.vae.decode(latents, return_dict=False)[0]
         image = (image / 2 + 0.5).clamp(0, 1)
         # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
@@ -728,15 +728,16 @@ def __call__(
                     t,
                     encoder_hidden_states=prompt_embeds,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
-                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
@@ -745,7 +746,7 @@ def __call__(
                         callback(i, t, latents)
 
         if not output_type == "latent":
-            image = self.vae.decode(latents / self.vae.config.scaling_factor).sample
+            image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]
             image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype)
         else:
             image = latents

Diff for: src/diffusers/pipelines/deepfloyd_if/pipeline_if_img2img.py

@@ -918,7 +918,8 @@ def __call__(
                     t,
                     encoder_hidden_states=prompt_embeds,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
@@ -930,8 +931,8 @@ def __call__(
 
                 # compute the previous noisy sample x_t -> x_t-1
                 intermediate_images = self.scheduler.step(
-                    noise_pred, t, intermediate_images, **extra_step_kwargs
-                ).prev_sample
+                    noise_pred, t, intermediate_images, **extra_step_kwargs, return_dict=False
+                )[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):

Diff for: src/diffusers/pipelines/deepfloyd_if/pipeline_if_img2img_superresolution.py

@@ -1036,7 +1036,8 @@ def __call__(
                     encoder_hidden_states=prompt_embeds,
                     class_labels=noise_level,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
@@ -1048,8 +1049,8 @@ def __call__(
 
                 # compute the previous noisy sample x_t -> x_t-1
                 intermediate_images = self.scheduler.step(
-                    noise_pred, t, intermediate_images, **extra_step_kwargs
-                ).prev_sample
+                    noise_pred, t, intermediate_images, **extra_step_kwargs, return_dict=False
+                )[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):

Diff for: src/diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting.py

@@ -1033,7 +1033,8 @@ def __call__(
                     t,
                     encoder_hidden_states=prompt_embeds,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
@@ -1047,8 +1048,8 @@ def __call__(
                 prev_intermediate_images = intermediate_images
 
                 intermediate_images = self.scheduler.step(
-                    noise_pred, t, intermediate_images, **extra_step_kwargs
-                ).prev_sample
+                    noise_pred, t, intermediate_images, **extra_step_kwargs, return_dict=False
+                )[0]
 
                 intermediate_images = (1 - mask_image) * prev_intermediate_images + mask_image * intermediate_images
 

Diff for: src/diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting_superresolution.py

@@ -1143,7 +1143,8 @@ def __call__(
                     encoder_hidden_states=prompt_embeds,
                     class_labels=noise_level,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
@@ -1157,8 +1158,8 @@ def __call__(
                 prev_intermediate_images = intermediate_images
 
                 intermediate_images = self.scheduler.step(
-                    noise_pred, t, intermediate_images, **extra_step_kwargs
-                ).prev_sample
+                    noise_pred, t, intermediate_images, **extra_step_kwargs, return_dict=False
+                )[0]
 
                 intermediate_images = (1 - mask_image) * prev_intermediate_images + mask_image * intermediate_images
 

Diff for: src/diffusers/pipelines/deepfloyd_if/pipeline_if_superresolution.py

@@ -886,7 +886,8 @@ def __call__(
                     encoder_hidden_states=prompt_embeds,
                     class_labels=noise_level,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
@@ -898,8 +899,8 @@ def __call__(
 
                 # compute the previous noisy sample x_t -> x_t-1
                 intermediate_images = self.scheduler.step(
-                    noise_pred, t, intermediate_images, **extra_step_kwargs
-                ).prev_sample
+                    noise_pred, t, intermediate_images, **extra_step_kwargs, return_dict=False
+                )[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):

Diff for: src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_controlnet.py

@@ -20,6 +20,7 @@
 import numpy as np
 import PIL.Image
 import torch
+import torch.nn.functional as F
 from torch import nn
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
@@ -579,9 +580,20 @@ def check_inputs(
                 )
 
         # Check `image`
-        if isinstance(self.controlnet, ControlNetModel):
+        is_compiled = hasattr(F, "scaled_dot_product_attention") and isinstance(
+            self.controlnet, torch._dynamo.eval_frame.OptimizedModule
+        )
+        if (
+            isinstance(self.controlnet, ControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, ControlNetModel)
+        ):
             self.check_image(image, prompt, prompt_embeds)
-        elif isinstance(self.controlnet, MultiControlNetModel):
+        elif (
+            isinstance(self.controlnet, MultiControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
+        ):
             if not isinstance(image, list):
                 raise TypeError("For multiple controlnets: `image` must be type `list`")
 
@@ -600,10 +612,18 @@ def check_inputs(
             assert False
 
         # Check `controlnet_conditioning_scale`
-        if isinstance(self.controlnet, ControlNetModel):
+        if (
+            isinstance(self.controlnet, ControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, ControlNetModel)
+        ):
             if not isinstance(controlnet_conditioning_scale, float):
                 raise TypeError("For single controlnet: `controlnet_conditioning_scale` must be type `float`.")
-        elif isinstance(self.controlnet, MultiControlNetModel):
+        elif (
+            isinstance(self.controlnet, MultiControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
+        ):
             if isinstance(controlnet_conditioning_scale, list):
                 if any(isinstance(i, list) for i in controlnet_conditioning_scale):
                     raise ValueError("A single batch of multiple conditionings are supported at the moment.")
@@ -910,7 +930,14 @@ def __call__(
         )
 
         # 4. Prepare image
-        if isinstance(self.controlnet, ControlNetModel):
+        is_compiled = hasattr(F, "scaled_dot_product_attention") and isinstance(
+            self.controlnet, torch._dynamo.eval_frame.OptimizedModule
+        )
+        if (
+            isinstance(self.controlnet, ControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, ControlNetModel)
+        ):
             image = self.prepare_image(
                 image=image,
                 width=width,
@@ -922,7 +949,11 @@ def __call__(
                 do_classifier_free_guidance=do_classifier_free_guidance,
                 guess_mode=guess_mode,
             )
-        elif isinstance(self.controlnet, MultiControlNetModel):
+        elif (
+            isinstance(self.controlnet, MultiControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
+        ):
             images = []
 
             for image_ in image:
@@ -1006,15 +1037,16 @@ def __call__(
                     cross_attention_kwargs=cross_attention_kwargs,
                     down_block_additional_residuals=down_block_res_samples,
                     mid_block_additional_residual=mid_block_res_sample,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
-                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):

Diff for: src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py

@@ -677,15 +677,17 @@ def __call__(
                 latent_model_input = torch.cat([latent_model_input, depth_mask], dim=1)
 
                 # predict the noise residual
-                noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds).sample
+                noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds, return_dict=False)[
+                    0
+                ]
 
                 # perform guidance
                 if do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
-                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):

Diff for: src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -462,7 +462,7 @@ def decode_latents(self, latents):
             FutureWarning,
         )
         latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents).sample
+        image = self.vae.decode(latents, return_dict=False)[0]
         image = (image / 2 + 0.5).clamp(0, 1)
         # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
@@ -734,15 +734,16 @@ def __call__(
                     t,
                     encoder_hidden_states=prompt_embeds,
                     cross_attention_kwargs=cross_attention_kwargs,
-                ).sample
+                    return_dict=False,
+                )[0]
 
                 # perform guidance
                 if do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
-                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
@@ -751,7 +752,7 @@ def __call__(
                         callback(i, t, latents)
 
         if not output_type == "latent":
-            image = self.vae.decode(latents / self.vae.config.scaling_factor).sample
+            image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]
             image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype)
         else:
             image = latents

Diff for: src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -878,15 +878,17 @@ def __call__(
                 latent_model_input = torch.cat([latent_model_input, mask, masked_image_latents], dim=1)
 
                 # predict the noise residual
-                noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds).sample
+                noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds, return_dict=False)[
+                    0
+                ]
 
                 # perform guidance
                 if do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
-                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):

Diff for: src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint_legacy.py

@@ -690,15 +690,17 @@ def __call__(
                 latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
 
                 # predict the noise residual
-                noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds).sample
+                noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds, return_dict=False)[
+                    0
+                ]
 
                 # perform guidance
                 if do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
-                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
                 # masking
                 if add_predicted_noise:
                     init_latents_proper = self.scheduler.add_noise(