328 lines
12 KiB
Python
328 lines
12 KiB
Python
"""
|
|
Adapted from https://github.com/huggingface/diffusers/blob/v0.30.3/src/diffusers/schedulers/scheduling_flow_match_euler_discrete.py.
|
|
"""
|
|
|
|
import math
|
|
from dataclasses import dataclass
|
|
from typing import List, Optional, Tuple, Union
|
|
|
|
import numpy as np
|
|
import torch
|
|
from diffusers.configuration_utils import ConfigMixin, register_to_config
|
|
from diffusers.schedulers.scheduling_utils import SchedulerMixin
|
|
from diffusers.utils import BaseOutput, logging
|
|
from torch.distributions import LogisticNormal
|
|
|
|
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
|
|
|
|
|
|
# TODO: may move to training_utils.py
|
|
def compute_density_for_timestep_sampling(
|
|
weighting_scheme: str,
|
|
batch_size: int,
|
|
logit_mean: float = 0.0,
|
|
logit_std: float = 1.0,
|
|
mode_scale: float = None,
|
|
):
|
|
if weighting_scheme == "logit_normal":
|
|
# See 3.1 in the SD3 paper ($rf/lognorm(0.00,1.00)$).
|
|
u = torch.normal(
|
|
mean=logit_mean, std=logit_std, size=(batch_size,), device="cpu"
|
|
)
|
|
u = torch.nn.functional.sigmoid(u)
|
|
elif weighting_scheme == "logit_normal_dist":
|
|
u = (
|
|
LogisticNormal(loc=logit_mean, scale=logit_std)
|
|
.sample((batch_size,))[:, 0]
|
|
.to("cpu")
|
|
)
|
|
elif weighting_scheme == "mode":
|
|
u = torch.rand(size=(batch_size,), device="cpu")
|
|
u = 1 - u - mode_scale * (torch.cos(math.pi * u / 2) ** 2 - 1 + u)
|
|
else:
|
|
u = torch.rand(size=(batch_size,), device="cpu")
|
|
return u
|
|
|
|
|
|
def compute_loss_weighting(weighting_scheme: str, sigmas=None):
|
|
"""
|
|
Computes loss weighting scheme for SD3 training.
|
|
|
|
Courtesy: This was contributed by Rafie Walker in https://github.com/huggingface/diffusers/pull/8528.
|
|
|
|
SD3 paper reference: https://arxiv.org/abs/2403.03206v1.
|
|
"""
|
|
if weighting_scheme == "sigma_sqrt":
|
|
weighting = (sigmas**-2.0).float()
|
|
elif weighting_scheme == "cosmap":
|
|
bot = 1 - 2 * sigmas + 2 * sigmas**2
|
|
weighting = 2 / (math.pi * bot)
|
|
else:
|
|
weighting = torch.ones_like(sigmas)
|
|
return weighting
|
|
|
|
|
|
@dataclass
|
|
class RectifiedFlowSchedulerOutput(BaseOutput):
|
|
"""
|
|
Output class for the scheduler's `step` function output.
|
|
|
|
Args:
|
|
prev_sample (`torch.FloatTensor` 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.
|
|
"""
|
|
|
|
prev_sample: torch.FloatTensor
|
|
|
|
|
|
class RectifiedFlowScheduler(SchedulerMixin, ConfigMixin):
|
|
"""
|
|
The rectified flow scheduler is a scheduler that is used to propagate the diffusion process in the rectified flow.
|
|
|
|
This model inherits from [`SchedulerMixin`] and [`ConfigMixin`]. Check the superclass documentation for the generic
|
|
methods the library implements for all schedulers such as loading and saving.
|
|
|
|
Args:
|
|
num_train_timesteps (`int`, defaults to 1000):
|
|
The number of diffusion steps to train the model.
|
|
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.
|
|
shift (`float`, defaults to 1.0):
|
|
The shift value for the timestep schedule.
|
|
"""
|
|
|
|
_compatibles = []
|
|
order = 1
|
|
|
|
@register_to_config
|
|
def __init__(
|
|
self,
|
|
num_train_timesteps: int = 1000,
|
|
shift: float = 1.0,
|
|
use_dynamic_shifting: bool = False,
|
|
):
|
|
# pre-compute timesteps and sigmas; no use in fact
|
|
# NOTE that shape diffusion sample timesteps randomly or in a distribution,
|
|
# instead of sampling from the pre-defined linspace
|
|
timesteps = np.array(
|
|
[
|
|
(1.0 - i / num_train_timesteps) * num_train_timesteps
|
|
for i in range(num_train_timesteps)
|
|
]
|
|
)
|
|
timesteps = torch.from_numpy(timesteps).to(dtype=torch.float32)
|
|
|
|
sigmas = timesteps / num_train_timesteps
|
|
if not use_dynamic_shifting:
|
|
# when use_dynamic_shifting is True, we apply the timestep shifting on the fly based on the image resolution
|
|
sigmas = self.time_shift(sigmas)
|
|
|
|
self.timesteps = sigmas * num_train_timesteps
|
|
|
|
self._step_index = None
|
|
self._begin_index = None
|
|
|
|
self.sigmas = sigmas.to("cpu") # to avoid too much CPU/GPU communication
|
|
|
|
@property
|
|
def step_index(self):
|
|
"""
|
|
The index counter for current timestep. It will increase 1 after each scheduler step.
|
|
"""
|
|
return self._step_index
|
|
|
|
@property
|
|
def begin_index(self):
|
|
"""
|
|
The index for the first timestep. It should be set from pipeline with `set_begin_index` method.
|
|
"""
|
|
return self._begin_index
|
|
|
|
# Copied from diffusers.schedulers.scheduling_dpmsolver_multistep.DPMSolverMultistepScheduler.set_begin_index
|
|
def set_begin_index(self, begin_index: int = 0):
|
|
"""
|
|
Sets the begin index for the scheduler. This function should be run from pipeline before the inference.
|
|
|
|
Args:
|
|
begin_index (`int`):
|
|
The begin index for the scheduler.
|
|
"""
|
|
self._begin_index = begin_index
|
|
|
|
def _sigma_to_t(self, sigma):
|
|
return sigma * self.config.num_train_timesteps
|
|
|
|
def _t_to_sigma(self, timestep):
|
|
return timestep / self.config.num_train_timesteps
|
|
|
|
def time_shift_dynamic(self, mu: float, sigma: float, t: torch.Tensor):
|
|
return math.exp(mu) / (math.exp(mu) + (1 / t - 1) ** sigma)
|
|
|
|
def time_shift(self, t: torch.Tensor):
|
|
return self.config.shift * t / (1 + (self.config.shift - 1) * t)
|
|
|
|
def set_timesteps(
|
|
self,
|
|
num_inference_steps: int = None,
|
|
device: Union[str, torch.device] = None,
|
|
sigmas: Optional[List[float]] = None,
|
|
mu: Optional[float] = None,
|
|
):
|
|
"""
|
|
Sets the discrete timesteps used for the diffusion chain (to be run before inference).
|
|
|
|
Args:
|
|
num_inference_steps (`int`):
|
|
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.
|
|
"""
|
|
|
|
if self.config.use_dynamic_shifting and mu is None:
|
|
raise ValueError(
|
|
" you have to pass a value for `mu` when `use_dynamic_shifting` is set to be `True`"
|
|
)
|
|
|
|
if sigmas is None:
|
|
self.num_inference_steps = num_inference_steps
|
|
timesteps = np.array(
|
|
[
|
|
(1.0 - i / num_inference_steps) * self.config.num_train_timesteps
|
|
for i in range(num_inference_steps)
|
|
]
|
|
) # different from the original code in SD3
|
|
sigmas = timesteps / self.config.num_train_timesteps
|
|
|
|
if self.config.use_dynamic_shifting:
|
|
sigmas = self.time_shift_dynamic(mu, 1.0, sigmas)
|
|
else:
|
|
sigmas = self.config.shift * sigmas / (1 + (self.config.shift - 1) * sigmas)
|
|
|
|
sigmas = torch.from_numpy(sigmas).to(dtype=torch.float32, device=device)
|
|
timesteps = sigmas * self.config.num_train_timesteps
|
|
|
|
self.timesteps = timesteps.to(device=device)
|
|
self.sigmas = torch.cat([sigmas, torch.zeros(1, device=sigmas.device)])
|
|
|
|
self._step_index = None
|
|
self._begin_index = None
|
|
|
|
def index_for_timestep(self, timestep, schedule_timesteps=None):
|
|
if schedule_timesteps is None:
|
|
schedule_timesteps = self.timesteps
|
|
|
|
indices = (schedule_timesteps == timestep).nonzero()
|
|
|
|
# The sigma index that is taken for the **very** first `step`
|
|
# is always the second index (or the last index if there is only 1)
|
|
# This way we can ensure we don't accidentally skip a sigma in
|
|
# case we start in the middle of the denoising schedule (e.g. for image-to-image)
|
|
pos = 1 if len(indices) > 1 else 0
|
|
|
|
return indices[pos].item()
|
|
|
|
def _init_step_index(self, timestep):
|
|
if self.begin_index is None:
|
|
if isinstance(timestep, torch.Tensor):
|
|
timestep = timestep.to(self.timesteps.device)
|
|
self._step_index = self.index_for_timestep(timestep)
|
|
else:
|
|
self._step_index = self._begin_index
|
|
|
|
def step(
|
|
self,
|
|
model_output: torch.FloatTensor,
|
|
timestep: Union[float, torch.FloatTensor],
|
|
sample: torch.FloatTensor,
|
|
s_churn: float = 0.0,
|
|
s_tmin: float = 0.0,
|
|
s_tmax: float = float("inf"),
|
|
s_noise: float = 1.0,
|
|
generator: Optional[torch.Generator] = None,
|
|
return_dict: bool = True,
|
|
) -> Union[RectifiedFlowSchedulerOutput, 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`):
|
|
The direct output from learned diffusion model.
|
|
timestep (`float`):
|
|
The current discrete timestep in the diffusion chain.
|
|
sample (`torch.FloatTensor`):
|
|
A current instance of a sample created by the diffusion process.
|
|
s_churn (`float`):
|
|
s_tmin (`float`):
|
|
s_tmax (`float`):
|
|
s_noise (`float`, defaults to 1.0):
|
|
Scaling factor for noise added to the sample.
|
|
generator (`torch.Generator`, *optional*):
|
|
A random number generator.
|
|
return_dict (`bool`):
|
|
Whether or not to return a [`~schedulers.scheduling_euler_discrete.EulerDiscreteSchedulerOutput`] or
|
|
tuple.
|
|
|
|
Returns:
|
|
[`~schedulers.scheduling_euler_discrete.EulerDiscreteSchedulerOutput`] or `tuple`:
|
|
If return_dict is `True`, [`~schedulers.scheduling_euler_discrete.EulerDiscreteSchedulerOutput`] is
|
|
returned, otherwise a tuple is returned where the first element is the sample tensor.
|
|
"""
|
|
|
|
if (
|
|
isinstance(timestep, int)
|
|
or isinstance(timestep, torch.IntTensor)
|
|
or isinstance(timestep, torch.LongTensor)
|
|
):
|
|
raise ValueError(
|
|
(
|
|
"Passing integer indices (e.g. from `enumerate(timesteps)`) as timesteps to"
|
|
" `EulerDiscreteScheduler.step()` is not supported. Make sure to pass"
|
|
" one of the `scheduler.timesteps` as a timestep."
|
|
),
|
|
)
|
|
|
|
if self.step_index is None:
|
|
self._init_step_index(timestep)
|
|
|
|
# Upcast to avoid precision issues when computing prev_sample
|
|
sample = sample.to(torch.float32)
|
|
|
|
sigma = self.sigmas[self.step_index]
|
|
sigma_next = self.sigmas[self.step_index + 1]
|
|
|
|
# Here different directions are used for the flow matching
|
|
prev_sample = sample + (sigma - sigma_next) * model_output
|
|
|
|
# Cast sample back to model compatible dtype
|
|
prev_sample = prev_sample.to(model_output.dtype)
|
|
|
|
# upon completion increase step index by one
|
|
self._step_index += 1
|
|
|
|
if not return_dict:
|
|
return (prev_sample,)
|
|
|
|
return RectifiedFlowSchedulerOutput(prev_sample=prev_sample)
|
|
|
|
def scale_noise(
|
|
self,
|
|
original_samples: torch.Tensor,
|
|
noise: torch.Tensor,
|
|
timesteps: torch.IntTensor,
|
|
) -> torch.Tensor:
|
|
"""
|
|
Forward function for the noise scaling in the flow matching.
|
|
"""
|
|
sigmas = self._t_to_sigma(timesteps.to(dtype=torch.float32))
|
|
|
|
while len(sigmas.shape) < len(original_samples.shape):
|
|
sigmas = sigmas.unsqueeze(-1)
|
|
|
|
return (1.0 - sigmas) * original_samples + sigmas * noise
|
|
|
|
def __len__(self):
|
|
return self.config.num_train_timesteps
|