Merge branch 'main' into redux

Author: linoytsaban

docs/source/en/conceptual/evaluation.md

@@ -181,7 +181,7 @@ Then we load the [v1-5 checkpoint](https://huggingface.co/stable-diffusion-v1-5/
 
 ```python
 model_ckpt_1_5 = "stable-diffusion-v1-5/stable-diffusion-v1-5"
-sd_pipeline_1_5 = StableDiffusionPipeline.from_pretrained(model_ckpt_1_5, torch_dtype=weight_dtype).to(device)
+sd_pipeline_1_5 = StableDiffusionPipeline.from_pretrained(model_ckpt_1_5, torch_dtype=torch.float16).to("cuda")
 
 images_1_5 = sd_pipeline_1_5(prompts, num_images_per_prompt=1, generator=generator, output_type="np").images
 ```
@@ -280,7 +280,7 @@ from diffusers import StableDiffusionInstructPix2PixPipeline
 
 instruct_pix2pix_pipeline = StableDiffusionInstructPix2PixPipeline.from_pretrained(
     "timbrooks/instruct-pix2pix", torch_dtype=torch.float16
-).to(device)
+).to("cuda")
 ```
 
 Now, we perform the edits:
@@ -326,9 +326,9 @@ from transformers import (
 
 clip_id = "openai/clip-vit-large-patch14"
 tokenizer = CLIPTokenizer.from_pretrained(clip_id)
-text_encoder = CLIPTextModelWithProjection.from_pretrained(clip_id).to(device)
+text_encoder = CLIPTextModelWithProjection.from_pretrained(clip_id).to("cuda")
 image_processor = CLIPImageProcessor.from_pretrained(clip_id)
-image_encoder = CLIPVisionModelWithProjection.from_pretrained(clip_id).to(device)
+image_encoder = CLIPVisionModelWithProjection.from_pretrained(clip_id).to("cuda")
 ```
 
 Notice that we are using a particular CLIP checkpoint, i.e., `openai/clip-vit-large-patch14`. This is because the Stable Diffusion pre-training was performed with this CLIP variant. For more details, refer to the [documentation](https://huggingface.co/docs/transformers/model_doc/clip).
@@ -350,7 +350,7 @@ class DirectionalSimilarity(nn.Module):
 
     def preprocess_image(self, image):
         image = self.image_processor(image, return_tensors="pt")["pixel_values"]
-        return {"pixel_values": image.to(device)}
+        return {"pixel_values": image.to("cuda")}
 
     def tokenize_text(self, text):
         inputs = self.tokenizer(
@@ -360,7 +360,7 @@ class DirectionalSimilarity(nn.Module):
             truncation=True,
             return_tensors="pt",
         )
-        return {"input_ids": inputs.input_ids.to(device)}
+        return {"input_ids": inputs.input_ids.to("cuda")}
 
     def encode_image(self, image):
         preprocessed_image = self.preprocess_image(image)
@@ -459,6 +459,7 @@ with ZipFile(local_filepath, "r") as zipper:
 ```python
 from PIL import Image
 import os
+import numpy as np
 
 dataset_path = "sample-imagenet-images"
 image_paths = sorted([os.path.join(dataset_path, x) for x in os.listdir(dataset_path)])
@@ -477,6 +478,7 @@ Now that the images are loaded, let's apply some lightweight pre-processing on t
 
 ```python
 from torchvision.transforms import functional as F
+import torch
 
 
 def preprocess_image(image):
@@ -498,6 +500,10 @@ dit_pipeline = DiTPipeline.from_pretrained("facebook/DiT-XL-2-256", torch_dtype=
 dit_pipeline.scheduler = DPMSolverMultistepScheduler.from_config(dit_pipeline.scheduler.config)
 dit_pipeline = dit_pipeline.to("cuda")
 
+seed = 0
+generator = torch.manual_seed(seed)
+
+
 words = [
     "cassette player",
     "chainsaw",

src/diffusers/pipelines/flux/pipeline_flux.py

@@ -554,7 +554,7 @@ def __call__(
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 28,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 3.5,
         num_images_per_prompt: Optional[int] = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
@@ -585,10 +585,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
-                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
-                passed will be used. Must be in descending order.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                 `guidance_scale` is defined as `w` of equation 2. of [Imagen
@@ -699,7 +699,7 @@ def __call__(
         )
 
         # 5. Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = latents.shape[1]
         mu = calculate_shift(
             image_seq_len,
@@ -712,8 +712,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
         num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)

src/diffusers/pipelines/flux/pipeline_flux_control.py

@@ -621,7 +621,7 @@ def __call__(
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 28,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 3.5,
         num_images_per_prompt: Optional[int] = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
@@ -660,10 +660,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
-                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
-                passed will be used. Must be in descending order.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                 `guidance_scale` is defined as `w` of equation 2. of [Imagen
@@ -799,7 +799,7 @@ def __call__(
         )
 
         # 5. Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = latents.shape[1]
         mu = calculate_shift(
             image_seq_len,
@@ -812,8 +812,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
         num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)

src/diffusers/pipelines/flux/pipeline_flux_control_img2img.py

@@ -647,7 +647,7 @@ def __call__(
         width: Optional[int] = None,
         strength: float = 0.6,
         num_inference_steps: int = 28,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 7.0,
         num_images_per_prompt: Optional[int] = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
@@ -698,10 +698,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
-                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
-                passed will be used. Must be in descending order.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                 `guidance_scale` is defined as `w` of equation 2. of [Imagen
@@ -805,7 +805,7 @@ def __call__(
         )
 
         # 4.Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = (int(height) // self.vae_scale_factor // 2) * (int(width) // self.vae_scale_factor // 2)
         mu = calculate_shift(
             image_seq_len,
@@ -818,8 +818,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
         timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, strength, device)

src/diffusers/pipelines/flux/pipeline_flux_controlnet.py

@@ -602,7 +602,7 @@ def __call__(
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 28,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 7.0,
         control_guidance_start: Union[float, List[float]] = 0.0,
         control_guidance_end: Union[float, List[float]] = 1.0,
@@ -638,10 +638,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
-                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
-                passed will be used. Must be in descending order.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                 `guidance_scale` is defined as `w` of equation 2. of [Imagen
@@ -872,7 +872,7 @@ def __call__(
         )
 
         # 5. Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = latents.shape[1]
         mu = calculate_shift(
             image_seq_len,
@@ -885,8 +885,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
 

src/diffusers/pipelines/flux/pipeline_flux_controlnet_image_to_image.py

@@ -646,7 +646,7 @@ def __call__(
         width: Optional[int] = None,
         strength: float = 0.6,
         num_inference_steps: int = 28,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 7.0,
         control_guidance_start: Union[float, List[float]] = 0.0,
         control_guidance_end: Union[float, List[float]] = 1.0,
@@ -685,8 +685,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 28):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
             control_mode (`int` or `List[int]`, *optional*):
@@ -858,7 +860,7 @@ def __call__(
             control_mode = torch.tensor(control_mode_).to(device, dtype=torch.long)
             control_mode = control_mode.reshape([-1, 1])
 
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = (int(height) // self.vae_scale_factor // 2) * (int(width) // self.vae_scale_factor // 2)
         mu = calculate_shift(
             image_seq_len,
@@ -871,8 +873,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
         timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, strength, device)

src/diffusers/pipelines/flux/pipeline_flux_controlnet_inpainting.py

@@ -752,7 +752,7 @@ def __call__(
         width: Optional[int] = None,
         strength: float = 0.6,
         padding_mask_crop: Optional[int] = None,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         num_inference_steps: int = 28,
         guidance_scale: float = 7.0,
         control_guidance_start: Union[float, List[float]] = 0.0,
@@ -799,8 +799,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 28):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
             control_guidance_start (`float` or `List[float]`, *optional*, defaults to 0.0):
@@ -1009,7 +1011,7 @@ def __call__(
 
         # 6. Prepare timesteps
 
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = (int(global_height) // self.vae_scale_factor // 2) * (
             int(global_width) // self.vae_scale_factor // 2
         )
@@ -1024,8 +1026,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
         timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, strength, device)

src/diffusers/pipelines/flux/pipeline_flux_fill.py

@@ -689,7 +689,7 @@ def __call__(
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 50,
-        timesteps: List[int] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 30.0,
         num_images_per_prompt: Optional[int] = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
@@ -735,10 +735,10 @@ def __call__(
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
-                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
-                passed will be used. Must be in descending order.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             guidance_scale (`float`, *optional*, defaults to 7.0):
                 Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                 `guidance_scale` is defined as `w` of equation 2. of [Imagen
@@ -878,7 +878,7 @@ def __call__(
             masked_image_latents = torch.cat((masked_image_latents, mask), dim=-1)
 
         # 6. Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
         image_seq_len = latents.shape[1]
         mu = calculate_shift(
             image_seq_len,
@@ -891,8 +891,7 @@ def __call__(
             self.scheduler,
             num_inference_steps,
             device,
-            timesteps,
-            sigmas,
+            sigmas=sigmas,
             mu=mu,
         )
         num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)