diffusers 0.27.1__py3-none-any.whl → 0.32.2__py3-none-any.whl

diffusers/schedulers/scheduling_heun_discrete.py

@@ -13,13 +13,38 @@
 # limitations under the License.
 
 import math
+from dataclasses import dataclass
 from typing import List, Optional, Tuple, Union
 
 import numpy as np
 import torch
 
 from ..configuration_utils import ConfigMixin, register_to_config
-from .scheduling_utils import KarrasDiffusionSchedulers, SchedulerMixin, SchedulerOutput
+from ..utils import BaseOutput, is_scipy_available
+from .scheduling_utils import KarrasDiffusionSchedulers, SchedulerMixin
+
+
+if is_scipy_available():
+    import scipy.stats
+
+
+@dataclass
+# Copied from diffusers.schedulers.scheduling_ddpm.DDPMSchedulerOutput with DDPM->HeunDiscrete
+class HeunDiscreteSchedulerOutput(BaseOutput):
+    """
+    Output class for the scheduler's `step` function output.
+
+    Args:
+        prev_sample (`torch.Tensor` of shape `(batch_size, num_channels, height, width)` for images):
+            Computed sample `(x_{t-1})` of previous timestep. `prev_sample` should be used as next model input in the
+            denoising loop.
+        pred_original_sample (`torch.Tensor` of shape `(batch_size, num_channels, height, width)` for images):
+            The predicted denoised sample `(x_{0})` based on the model output from the current timestep.
+            `pred_original_sample` can be used to preview progress or for guidance.
+    """
+
+    prev_sample: torch.Tensor
+    pred_original_sample: Optional[torch.Tensor] = None
 
 
 # Copied from diffusers.schedulers.scheduling_ddpm.betas_for_alpha_bar
@@ -57,7 +82,7 @@ def betas_for_alpha_bar(
             return math.exp(t * -12.0)
 
     else:
-        raise ValueError(f"Unsupported alpha_tranform_type: {alpha_transform_type}")
+        raise ValueError(f"Unsupported alpha_transform_type: {alpha_transform_type}")
 
     betas = []
     for i in range(num_diffusion_timesteps):
@@ -97,6 +122,11 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         use_karras_sigmas (`bool`, *optional*, defaults to `False`):
             Whether to use Karras sigmas for step sizes in the noise schedule during the sampling process. If `True`,
             the sigmas are determined according to a sequence of noise levels {σi}.
+        use_exponential_sigmas (`bool`, *optional*, defaults to `False`):
+            Whether to use exponential sigmas for step sizes in the noise schedule during the sampling process.
+        use_beta_sigmas (`bool`, *optional*, defaults to `False`):
+            Whether to use beta sigmas for step sizes in the noise schedule during the sampling process. Refer to [Beta
+            Sampling is All You Need](https://huggingface.co/papers/2407.12173) for more information.
         timestep_spacing (`str`, defaults to `"linspace"`):
             The way the timesteps should be scaled. Refer to Table 2 of the [Common Diffusion Noise Schedules and
             Sample Steps are Flawed](https://huggingface.co/papers/2305.08891) for more information.
@@ -117,11 +147,19 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         trained_betas: Optional[Union[np.ndarray, List[float]]] = None,
         prediction_type: str = "epsilon",
         use_karras_sigmas: Optional[bool] = False,
+        use_exponential_sigmas: Optional[bool] = False,
+        use_beta_sigmas: Optional[bool] = False,
         clip_sample: Optional[bool] = False,
         clip_sample_range: float = 1.0,
         timestep_spacing: str = "linspace",
         steps_offset: int = 0,
     ):
+        if self.config.use_beta_sigmas and not is_scipy_available():
+            raise ImportError("Make sure to install scipy if you want to use beta sigmas.")
+        if sum([self.config.use_beta_sigmas, self.config.use_exponential_sigmas, self.config.use_karras_sigmas]) > 1:
+            raise ValueError(
+                "Only one of `config.use_beta_sigmas`, `config.use_exponential_sigmas`, `config.use_karras_sigmas` can be used."
+            )
         if trained_betas is not None:
             self.betas = torch.tensor(trained_betas, dtype=torch.float32)
         elif beta_schedule == "linear":
@@ -135,7 +173,7 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         elif beta_schedule == "exp":
             self.betas = betas_for_alpha_bar(num_train_timesteps, alpha_transform_type="exp")
         else:
-            raise NotImplementedError(f"{beta_schedule} does is not implemented for {self.__class__}")
+            raise NotImplementedError(f"{beta_schedule} is not implemented for {self.__class__}")
 
         self.alphas = 1.0 - self.betas
         self.alphas_cumprod = torch.cumprod(self.alphas, dim=0)
@@ -174,7 +212,7 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
     @property
     def step_index(self):
         """
-        The index counter for current timestep. It will increae 1 after each scheduler step.
+        The index counter for current timestep. It will increase 1 after each scheduler step.
         """
         return self._step_index
 
@@ -198,21 +236,21 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
 
     def scale_model_input(
         self,
-        sample: torch.FloatTensor,
-        timestep: Union[float, torch.FloatTensor],
-    ) -> torch.FloatTensor:
+        sample: torch.Tensor,
+        timestep: Union[float, torch.Tensor],
+    ) -> torch.Tensor:
         """
         Ensures interchangeability with schedulers that need to scale the denoising model input depending on the
         current timestep.
 
         Args:
-            sample (`torch.FloatTensor`):
+            sample (`torch.Tensor`):
                 The input sample.
             timestep (`int`, *optional*):
                 The current timestep in the diffusion chain.
 
         Returns:
-            `torch.FloatTensor`:
+            `torch.Tensor`:
                 A scaled input sample.
         """
         if self.step_index is None:
@@ -224,9 +262,10 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
 
     def set_timesteps(
         self,
-        num_inference_steps: int,
+        num_inference_steps: Optional[int] = None,
         device: Union[str, torch.device] = None,
         num_train_timesteps: Optional[int] = None,
+        timesteps: Optional[List[int]] = None,
     ):
         """
         Sets the discrete timesteps used for the diffusion chain (to be run before inference).
@@ -236,30 +275,51 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
                 The number of diffusion steps used when generating samples with a pre-trained model.
             device (`str` or `torch.device`, *optional*):
                 The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+            num_train_timesteps (`int`, *optional*):
+                The number of diffusion steps used when training the model. If `None`, the default
+                `num_train_timesteps` attribute is used.
+            timesteps (`List[int]`, *optional*):
+                Custom timesteps used to support arbitrary spacing between timesteps. If `None`, timesteps will be
+                generated based on the `timestep_spacing` attribute. If `timesteps` is passed, `num_inference_steps`
+                must be `None`, and `timestep_spacing` attribute will be ignored.
         """
+        if num_inference_steps is None and timesteps is None:
+            raise ValueError("Must pass exactly one of `num_inference_steps` or `custom_timesteps`.")
+        if num_inference_steps is not None and timesteps is not None:
+            raise ValueError("Can only pass one of `num_inference_steps` or `custom_timesteps`.")
+        if timesteps is not None and self.config.use_karras_sigmas:
+            raise ValueError("Cannot use `timesteps` with `config.use_karras_sigmas = True`")
+        if timesteps is not None and self.config.use_exponential_sigmas:
+            raise ValueError("Cannot set `timesteps` with `config.use_exponential_sigmas = True`.")
+        if timesteps is not None and self.config.use_beta_sigmas:
+            raise ValueError("Cannot set `timesteps` with `config.use_beta_sigmas = True`.")
+
+        num_inference_steps = num_inference_steps or len(timesteps)
         self.num_inference_steps = num_inference_steps
-
         num_train_timesteps = num_train_timesteps or self.config.num_train_timesteps
 
-        # "linspace", "leading", "trailing" corresponds to annotation of Table 2. of https://arxiv.org/abs/2305.08891
-        if self.config.timestep_spacing == "linspace":
-            timesteps = np.linspace(0, num_train_timesteps - 1, num_inference_steps, dtype=np.float32)[::-1].copy()
-        elif self.config.timestep_spacing == "leading":
-            step_ratio = num_train_timesteps // self.num_inference_steps
-            # creates integer timesteps by multiplying by ratio
-            # casting to int to avoid issues when num_inference_step is power of 3
-            timesteps = (np.arange(0, num_inference_steps) * step_ratio).round()[::-1].copy().astype(np.float32)
-            timesteps += self.config.steps_offset
-        elif self.config.timestep_spacing == "trailing":
-            step_ratio = num_train_timesteps / self.num_inference_steps
-            # creates integer timesteps by multiplying by ratio
-            # casting to int to avoid issues when num_inference_step is power of 3
-            timesteps = (np.arange(num_train_timesteps, 0, -step_ratio)).round().copy().astype(np.float32)
-            timesteps -= 1
+        if timesteps is not None:
+            timesteps = np.array(timesteps, dtype=np.float32)
         else:
-            raise ValueError(
-                f"{self.config.timestep_spacing} is not supported. Please make sure to choose one of 'linspace', 'leading' or 'trailing'."
-            )
+            # "linspace", "leading", "trailing" corresponds to annotation of Table 2. of https://arxiv.org/abs/2305.08891
+            if self.config.timestep_spacing == "linspace":
+                timesteps = np.linspace(0, num_train_timesteps - 1, num_inference_steps, dtype=np.float32)[::-1].copy()
+            elif self.config.timestep_spacing == "leading":
+                step_ratio = num_train_timesteps // self.num_inference_steps
+                # creates integer timesteps by multiplying by ratio
+                # casting to int to avoid issues when num_inference_step is power of 3
+                timesteps = (np.arange(0, num_inference_steps) * step_ratio).round()[::-1].copy().astype(np.float32)
+                timesteps += self.config.steps_offset
+            elif self.config.timestep_spacing == "trailing":
+                step_ratio = num_train_timesteps / self.num_inference_steps
+                # creates integer timesteps by multiplying by ratio
+                # casting to int to avoid issues when num_inference_step is power of 3
+                timesteps = (np.arange(num_train_timesteps, 0, -step_ratio)).round().copy().astype(np.float32)
+                timesteps -= 1
+            else:
+                raise ValueError(
+                    f"{self.config.timestep_spacing} is not supported. Please make sure to choose one of 'linspace', 'leading' or 'trailing'."
+                )
 
         sigmas = np.array(((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5)
         log_sigmas = np.log(sigmas)
@@ -268,6 +328,12 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         if self.config.use_karras_sigmas:
             sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
             timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
+        elif self.config.use_exponential_sigmas:
+            sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
+            timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
+        elif self.config.use_beta_sigmas:
+            sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
+            timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
 
         sigmas = np.concatenate([sigmas, [0.0]]).astype(np.float32)
         sigmas = torch.from_numpy(sigmas).to(device=device)
@@ -311,7 +377,7 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         return t
 
     # Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_karras
-    def _convert_to_karras(self, in_sigmas: torch.FloatTensor, num_inference_steps) -> torch.FloatTensor:
+    def _convert_to_karras(self, in_sigmas: torch.Tensor, num_inference_steps) -> torch.Tensor:
         """Constructs the noise schedule of Karras et al. (2022)."""
 
         # Hack to make sure that other schedulers which copy this function don't break
@@ -336,6 +402,60 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         sigmas = (max_inv_rho + ramp * (min_inv_rho - max_inv_rho)) ** rho
         return sigmas
 
+    # Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_exponential
+    def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps: int) -> torch.Tensor:
+        """Constructs an exponential noise schedule."""
+
+        # Hack to make sure that other schedulers which copy this function don't break
+        # TODO: Add this logic to the other schedulers
+        if hasattr(self.config, "sigma_min"):
+            sigma_min = self.config.sigma_min
+        else:
+            sigma_min = None
+
+        if hasattr(self.config, "sigma_max"):
+            sigma_max = self.config.sigma_max
+        else:
+            sigma_max = None
+
+        sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
+        sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()
+
+        sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
+        return sigmas
+
+    # Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
+    def _convert_to_beta(
+        self, in_sigmas: torch.Tensor, num_inference_steps: int, alpha: float = 0.6, beta: float = 0.6
+    ) -> torch.Tensor:
+        """From "Beta Sampling is All You Need" [arXiv:2407.12173] (Lee et. al, 2024)"""
+
+        # Hack to make sure that other schedulers which copy this function don't break
+        # TODO: Add this logic to the other schedulers
+        if hasattr(self.config, "sigma_min"):
+            sigma_min = self.config.sigma_min
+        else:
+            sigma_min = None
+
+        if hasattr(self.config, "sigma_max"):
+            sigma_max = self.config.sigma_max
+        else:
+            sigma_max = None
+
+        sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
+        sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()
+
+        sigmas = np.array(
+            [
+                sigma_min + (ppf * (sigma_max - sigma_min))
+                for ppf in [
+                    scipy.stats.beta.ppf(timestep, alpha, beta)
+                    for timestep in 1 - np.linspace(0, 1, num_inference_steps)
+                ]
+            ]
+        )
+        return sigmas
+
     @property
     def state_in_first_order(self):
         return self.dt is None
@@ -351,29 +471,30 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
 
     def step(
         self,
-        model_output: Union[torch.FloatTensor, np.ndarray],
-        timestep: Union[float, torch.FloatTensor],
-        sample: Union[torch.FloatTensor, np.ndarray],
+        model_output: Union[torch.Tensor, np.ndarray],
+        timestep: Union[float, torch.Tensor],
+        sample: Union[torch.Tensor, np.ndarray],
         return_dict: bool = True,
-    ) -> Union[SchedulerOutput, Tuple]:
+    ) -> Union[HeunDiscreteSchedulerOutput, Tuple]:
         """
         Predict the sample from the previous timestep by reversing the SDE. This function propagates the diffusion
         process from the learned model outputs (most often the predicted noise).
 
         Args:
-            model_output (`torch.FloatTensor`):
+            model_output (`torch.Tensor`):
                 The direct output from learned diffusion model.
             timestep (`float`):
                 The current discrete timestep in the diffusion chain.
-            sample (`torch.FloatTensor`):
+            sample (`torch.Tensor`):
                 A current instance of a sample created by the diffusion process.
             return_dict (`bool`):
-                Whether or not to return a [`~schedulers.scheduling_utils.SchedulerOutput`] or tuple.
+                Whether or not to return a [`~schedulers.scheduling_heun_discrete.HeunDiscreteSchedulerOutput`] or
+                tuple.
 
         Returns:
-            [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`:
-                If return_dict is `True`, [`~schedulers.scheduling_utils.SchedulerOutput`] is returned, otherwise a
-                tuple is returned where the first element is the sample tensor.
+            [`~schedulers.scheduling_heun_discrete.HeunDiscreteSchedulerOutput`] or `tuple`:
+                If return_dict is `True`, [`~schedulers.scheduling_heun_discrete.HeunDiscreteSchedulerOutput`] is
+                returned, otherwise a tuple is returned where the first element is the sample tensor.
         """
         if self.step_index is None:
             self._init_step_index(timestep)
@@ -444,17 +565,20 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         self._step_index += 1
 
         if not return_dict:
-            return (prev_sample,)
+            return (
+                prev_sample,
+                pred_original_sample,
+            )
 
-        return SchedulerOutput(prev_sample=prev_sample)
+        return HeunDiscreteSchedulerOutput(prev_sample=prev_sample, pred_original_sample=pred_original_sample)
 
     # Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler.add_noise
     def add_noise(
         self,
-        original_samples: torch.FloatTensor,
-        noise: torch.FloatTensor,
-        timesteps: torch.FloatTensor,
-    ) -> torch.FloatTensor:
+        original_samples: torch.Tensor,
+        noise: torch.Tensor,
+        timesteps: torch.Tensor,
+    ) -> torch.Tensor:
         # Make sure sigmas and timesteps have the same device and dtype as original_samples
         sigmas = self.sigmas.to(device=original_samples.device, dtype=original_samples.dtype)
         if original_samples.device.type == "mps" and torch.is_floating_point(timesteps):
@@ -468,7 +592,11 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called before first denoising step to create initial latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

diffusers/schedulers/scheduling_ipndm.py

@@ -61,7 +61,7 @@ class IPNDMScheduler(SchedulerMixin, ConfigMixin):
     @property
     def step_index(self):
         """
-        The index counter for current timestep. It will increae 1 after each scheduler step.
+        The index counter for current timestep. It will increase 1 after each scheduler step.
         """
         return self._step_index
 
@@ -137,9 +137,9 @@ class IPNDMScheduler(SchedulerMixin, ConfigMixin):
 
     def step(
         self,
-        model_output: torch.FloatTensor,
-        timestep: int,
-        sample: torch.FloatTensor,
+        model_output: torch.Tensor,
+        timestep: Union[int, torch.Tensor],
+        sample: torch.Tensor,
         return_dict: bool = True,
     ) -> Union[SchedulerOutput, Tuple]:
         """
@@ -147,11 +147,11 @@ class IPNDMScheduler(SchedulerMixin, ConfigMixin):
         the linear multistep method. It performs one forward pass multiple times to approximate the solution.
 
         Args:
-            model_output (`torch.FloatTensor`):
+            model_output (`torch.Tensor`):
                 The direct output from learned diffusion model.
             timestep (`int`):
                 The current discrete timestep in the diffusion chain.
-            sample (`torch.FloatTensor`):
+            sample (`torch.Tensor`):
                 A current instance of a sample created by the diffusion process.
             return_dict (`bool`):
                 Whether or not to return a [`~schedulers.scheduling_utils.SchedulerOutput`] or tuple.
@@ -193,17 +193,17 @@ class IPNDMScheduler(SchedulerMixin, ConfigMixin):
 
         return SchedulerOutput(prev_sample=prev_sample)
 
-    def scale_model_input(self, sample: torch.FloatTensor, *args, **kwargs) -> torch.FloatTensor:
+    def scale_model_input(self, sample: torch.Tensor, *args, **kwargs) -> torch.Tensor:
         """
         Ensures interchangeability with schedulers that need to scale the denoising model input depending on the
         current timestep.
 
         Args:
-            sample (`torch.FloatTensor`):
+            sample (`torch.Tensor`):
                 The input sample.
 
         Returns:
-            `torch.FloatTensor`:
+            `torch.Tensor`:
                 A scaled input sample.
         """
         return sample