multimodalart's picture
multimodalart HF Staff
Bernini-Diffusers-v2 r2v demo
fed6c68 verified
Raw
History Blame Contribute Delete
3.53 kB
# Copyright (c) 2026 Bytedance Ltd. and/or its affiliate
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Flow-matching scheduler for Wan sampling (inference only)."""
import torch
class FlowMatchScheduler:
def __init__(
self,
num_inference_steps: int = 100,
num_train_timesteps: int = 1000,
shift: float = 3.0,
sigma_max: float = 1.0,
sigma_min: float = 0.003 / 1.002,
inverse_timesteps: bool = False,
extra_one_step: bool = False,
reverse_sigmas: bool = False,
):
self.num_train_timesteps = num_train_timesteps
self.shift = shift
self.sigma_max = sigma_max
self.sigma_min = sigma_min
self.inverse_timesteps = inverse_timesteps
self.extra_one_step = extra_one_step
self.reverse_sigmas = reverse_sigmas
self.set_timesteps(num_inference_steps)
def set_timesteps(
self,
num_inference_steps: int = 100,
denoising_strength: float = 1.0,
shift: float = None,
device: str = None,
dtype: torch.dtype = torch.bfloat16,
training: bool = False,
):
if shift is not None:
self.shift = shift
if device is None:
device = "cuda" if torch.cuda.is_available() else "cpu"
sigma_start = self.sigma_min + (self.sigma_max - self.sigma_min) * denoising_strength
if self.extra_one_step:
self.sigmas = torch.linspace(
sigma_start, self.sigma_min, num_inference_steps + 1, device=device, dtype=dtype
)[:-1]
else:
self.sigmas = torch.linspace(
sigma_start, self.sigma_min, num_inference_steps, device=device, dtype=dtype
)
if self.inverse_timesteps:
self.sigmas = torch.flip(self.sigmas, dims=[0])
self.sigmas = self.shift * self.sigmas / (1 + (self.shift - 1) * self.sigmas)
if self.reverse_sigmas:
self.sigmas = 1 - self.sigmas
self.timesteps = self.sigmas * self.num_train_timesteps
self.training = training
def get_noise_sigma(self, timestep):
timestep = timestep.to(self.timesteps.device) if isinstance(timestep, torch.Tensor) else torch.tensor(timestep, device=self.timesteps.device)
timestep_id = torch.argmin((self.timesteps.unsqueeze(-1) - timestep.reshape(1, -1)).abs(), dim=0)
return self.sigmas[timestep_id].to(timestep.device)
def step(self, model_output, timestep, sample, to_final: bool = False, **kwargs):
if isinstance(timestep, torch.Tensor):
timestep = timestep.cuda(non_blocking=True)
timestep_id = torch.argmin((self.timesteps - timestep).abs())
sigma = self.sigmas[timestep_id]
if to_final or timestep_id + 1 >= len(self.timesteps):
sigma_ = 1 if (self.inverse_timesteps or self.reverse_sigmas) else 0
else:
sigma_ = self.sigmas[timestep_id + 1]
return sample + model_output * (sigma_ - sigma)