mirror of https://github.com/vladmandic/automatic
365 lines
18 KiB
Python
365 lines
18 KiB
Python
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, SchedulerOutput
|
|
|
|
|
|
# pylint: disable=no-member
|
|
class RadauIIAScheduler(SchedulerMixin, ConfigMixin):
|
|
"""
|
|
RadauIIAScheduler: Fully implicit Runge-Kutta integrators.
|
|
Supports variants with 2, 3, 5, 7, 9, 11 stages.
|
|
Adapted from the RES4LYF repository.
|
|
"""
|
|
|
|
order = 1
|
|
|
|
@register_to_config
|
|
def __init__(
|
|
self,
|
|
num_train_timesteps: int = 1000,
|
|
beta_start: float = 0.00085,
|
|
beta_end: float = 0.012,
|
|
beta_schedule: str = "linear",
|
|
trained_betas: Optional[Union[np.ndarray, List[float]]] = None,
|
|
prediction_type: str = "epsilon",
|
|
variant: str = "radau_iia_3s", # 2s to 11s variants
|
|
use_karras_sigmas: bool = False,
|
|
use_exponential_sigmas: bool = False,
|
|
use_beta_sigmas: bool = False,
|
|
use_flow_sigmas: bool = False,
|
|
sigma_min: Optional[float] = None,
|
|
sigma_max: Optional[float] = None,
|
|
rho: float = 7.0,
|
|
shift: Optional[float] = None,
|
|
base_shift: float = 0.5,
|
|
max_shift: float = 1.15,
|
|
use_dynamic_shifting: bool = False,
|
|
timestep_spacing: str = "linspace",
|
|
clip_sample: bool = False,
|
|
sample_max_value: float = 1.0,
|
|
set_alpha_to_one: bool = False,
|
|
skip_prk_steps: bool = False,
|
|
interpolation_type: str = "linear",
|
|
steps_offset: int = 0,
|
|
timestep_type: str = "discrete",
|
|
rescale_betas_zero_snr: bool = False,
|
|
final_sigmas_type: str = "zero",
|
|
):
|
|
if trained_betas is not None:
|
|
self.betas = torch.tensor(trained_betas, dtype=torch.float32)
|
|
elif beta_schedule == "linear":
|
|
self.betas = torch.linspace(beta_start, beta_end, num_train_timesteps, dtype=torch.float32)
|
|
elif beta_schedule == "scaled_linear":
|
|
self.betas = torch.linspace(beta_start**0.5, beta_end**0.5, num_train_timesteps, dtype=torch.float32) ** 2
|
|
else:
|
|
raise NotImplementedError(f"{beta_schedule} is not implemented")
|
|
|
|
self.alphas = 1.0 - self.betas
|
|
self.alphas_cumprod = torch.cumprod(self.alphas, dim=0)
|
|
|
|
self.num_inference_steps = None
|
|
self.timesteps = torch.from_numpy(np.arange(0, num_train_timesteps)[::-1].copy())
|
|
self.sigmas = None
|
|
self.init_noise_sigma = 1.0
|
|
|
|
# Internal state
|
|
self.model_outputs = []
|
|
self.sample_at_start_of_step = None
|
|
self._step_index = None
|
|
|
|
def _get_tableau(self):
|
|
v = self.config.variant
|
|
if v == "radau_iia_2s":
|
|
a, b, c = [[5 / 12, -1 / 12], [3 / 4, 1 / 4]], [3 / 4, 1 / 4], [1 / 3, 1]
|
|
elif v == "radau_iia_3s":
|
|
r6 = 6**0.5
|
|
a = [[11 / 45 - 7 * r6 / 360, 37 / 225 - 169 * r6 / 1800, -2 / 225 + r6 / 75], [37 / 225 + 169 * r6 / 1800, 11 / 45 + 7 * r6 / 360, -2 / 225 - r6 / 75], [4 / 9 - r6 / 36, 4 / 9 + r6 / 36, 1 / 9]]
|
|
b, c = [4 / 9 - r6 / 36, 4 / 9 + r6 / 36, 1 / 9], [2 / 5 - r6 / 10, 2 / 5 + r6 / 10, 1]
|
|
elif v == "radau_iia_5s":
|
|
a = [
|
|
[0.07299886, -0.02673533, 0.01867693, -0.01287911, 0.00504284],
|
|
[0.15377523, 0.14621487, -0.03644457, 0.02123306, -0.00793558],
|
|
[0.14006305, 0.29896713, 0.16758507, -0.03396910, 0.01094429],
|
|
[0.14489431, 0.27650007, 0.32579792, 0.12875675, -0.01570892],
|
|
[0.14371356, 0.28135602, 0.31182652, 0.22310390, 0.04000000],
|
|
]
|
|
b = [0.14371356, 0.28135602, 0.31182652, 0.22310390, 0.04]
|
|
c = [0.05710420, 0.27684301, 0.58359043, 0.86024014, 1.0]
|
|
elif v == "radau_iia_7s":
|
|
a = [
|
|
[0.03754626, -0.01403933, 0.01035279, -0.00815832, 0.00638841, -0.00460233, 0.00182894],
|
|
[0.08014760, 0.08106206, -0.02123799, 0.01400029, -0.01023419, 0.00715347, -0.00281264],
|
|
[0.07206385, 0.17106835, 0.10961456, -0.02461987, 0.01476038, -0.00957526, 0.00367268],
|
|
[0.07570513, 0.15409016, 0.22710774, 0.11747819, -0.02381083, 0.01270999, -0.00460884],
|
|
[0.07391234, 0.16135561, 0.20686724, 0.23700712, 0.10308679, -0.01885414, 0.00585890],
|
|
[0.07470556, 0.15830722, 0.21415342, 0.21987785, 0.19875212, 0.06926550, -0.00811601],
|
|
[0.07449424, 0.15910212, 0.21235189, 0.22355491, 0.19047494, 0.11961374, 0.02040816],
|
|
]
|
|
b = [0.07449424, 0.15910212, 0.21235189, 0.22355491, 0.19047494, 0.11961374, 0.02040816]
|
|
c = [0.02931643, 0.14807860, 0.33698469, 0.55867152, 0.76923386, 0.92694567, 1.0]
|
|
elif v == "radau_iia_9s":
|
|
a = [
|
|
[0.02278838, -0.00858964, 0.00645103, -0.00525753, 0.00438883, -0.00365122, 0.00294049, -0.00214927, 0.00085884],
|
|
[0.04890795, 0.05070205, -0.01352381, 0.00920937, -0.00715571, 0.00574725, -0.00454258, 0.00328816, -0.00130907],
|
|
[0.04374276, 0.10830189, 0.07291957, -0.01687988, 0.01070455, -0.00790195, 0.00599141, -0.00424802, 0.00167815],
|
|
[0.04624924, 0.09656073, 0.15429877, 0.08671937, -0.01845164, 0.01103666, -0.00767328, 0.00522822, -0.00203591],
|
|
[0.04483444, 0.10230685, 0.13821763, 0.18126393, 0.09043360, -0.01808506, 0.01019339, -0.00640527, 0.00242717],
|
|
[0.04565876, 0.09914547, 0.14574704, 0.16364828, 0.18594459, 0.08361326, -0.01580994, 0.00813825, -0.00291047],
|
|
[0.04520060, 0.10085371, 0.14194224, 0.17118947, 0.16978339, 0.16776829, 0.06707903, -0.01179223, 0.00360925],
|
|
[0.04541652, 0.10006040, 0.14365284, 0.16801908, 0.17556077, 0.15588627, 0.12889391, 0.04281083, -0.00493457],
|
|
[0.04535725, 0.10027665, 0.14319335, 0.16884698, 0.17413650, 0.15842189, 0.12359469, 0.07382701, 0.01234568],
|
|
]
|
|
b = [0.04535725, 0.10027665, 0.14319335, 0.16884698, 0.17413650, 0.15842189, 0.12359469, 0.07382701, 0.01234568]
|
|
c = [0.01777992, 0.09132361, 0.21430848, 0.37193216, 0.54518668, 0.71317524, 0.85563374, 0.95536604, 1.0]
|
|
elif v == "radau_iia_11s":
|
|
a = [
|
|
[0.01528052, -0.00578250, 0.00438010, -0.00362104, 0.00309298, -0.00267283, 0.00230509, -0.00195565, 0.00159387, -0.00117286, 0.00046993],
|
|
[0.03288398, 0.03451351, -0.00928542, 0.00641325, -0.00509546, 0.00424609, -0.00358767, 0.00300683, -0.00243267, 0.00178278, -0.00071315],
|
|
[0.02933250, 0.07416243, 0.05114868, -0.01200502, 0.00777795, -0.00594470, 0.00480266, -0.00392360, 0.00312733, -0.00227314, 0.00090638],
|
|
[0.03111455, 0.06578995, 0.10929963, 0.06381052, -0.01385359, 0.00855744, -0.00630764, 0.00491336, -0.00381400, 0.00273343, -0.00108397],
|
|
[0.03005269, 0.07011285, 0.09714692, 0.13539160, 0.07147108, -0.01471024, 0.00873319, -0.00619941, 0.00459164, -0.00321333, 0.00126286],
|
|
[0.03072807, 0.06751926, 0.10334060, 0.12083526, 0.15032679, 0.07350932, -0.01451288, 0.00829665, -0.00561283, 0.00376623, -0.00145771],
|
|
[0.03029202, 0.06914472, 0.09972096, 0.12801064, 0.13493180, 0.15289670, 0.06975993, -0.01327455, 0.00725877, -0.00448439, 0.00168785],
|
|
[0.03056654, 0.06813851, 0.10188107, 0.12403361, 0.14211432, 0.13829395, 0.14289135, 0.06052636, -0.01107774, 0.00559867, -0.00198773],
|
|
[0.03040663, 0.06871881, 0.10066096, 0.12619527, 0.13848876, 0.14450774, 0.13065189, 0.12111401, 0.04655548, -0.00802620, 0.00243764],
|
|
[0.03048412, 0.06843925, 0.10124185, 0.12518732, 0.14011843, 0.14190387, 0.13500343, 0.11262870, 0.08930604, 0.02896966, -0.00331170],
|
|
[0.03046255, 0.06851684, 0.10108155, 0.12546269, 0.13968067, 0.14258278, 0.13393354, 0.11443306, 0.08565881, 0.04992304, 0.00826446],
|
|
]
|
|
b = [0.03046255, 0.06851684, 0.10108155, 0.12546269, 0.13968067, 0.14258278, 0.13393354, 0.11443306, 0.08565881, 0.04992304, 0.00826446]
|
|
c = [0.01191761, 0.06173207, 0.14711145, 0.26115968, 0.39463985, 0.53673877, 0.67594446, 0.80097892, 0.90171099, 0.96997097, 1.0]
|
|
else:
|
|
raise ValueError(f"Unknown variant: {v}")
|
|
return np.array(a), np.array(b), np.array(c)
|
|
|
|
def set_timesteps(
|
|
self,
|
|
num_inference_steps: int,
|
|
device: Union[str, torch.device] = None,
|
|
mu: Optional[float] = None, dtype: torch.dtype = torch.float32):
|
|
self.num_inference_steps = num_inference_steps
|
|
|
|
# 1. Spacing
|
|
if self.config.timestep_spacing == "linspace":
|
|
timesteps = np.linspace(0, self.config.num_train_timesteps - 1, num_inference_steps, dtype=float)[::-1].copy()
|
|
elif self.config.timestep_spacing == "leading":
|
|
step_ratio = self.config.num_train_timesteps // num_inference_steps
|
|
timesteps = (np.arange(0, num_inference_steps) * step_ratio).round()[::-1].copy().astype(float)
|
|
elif self.config.timestep_spacing == "trailing":
|
|
step_ratio = self.config.num_train_timesteps / num_inference_steps
|
|
timesteps = (np.arange(self.config.num_train_timesteps, 0, -step_ratio)).round().copy().astype(float)
|
|
timesteps -= 1
|
|
else:
|
|
raise ValueError(f"timestep_spacing must be one of 'linspace', 'leading', or 'trailing', got {self.config.timestep_spacing}")
|
|
|
|
sigmas = np.array(((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5)
|
|
if self.config.interpolation_type == "linear":
|
|
sigmas = np.interp(timesteps, np.arange(len(sigmas)), sigmas)
|
|
elif self.config.interpolation_type == "log_linear":
|
|
sigmas = np.exp(np.interp(timesteps, np.arange(len(sigmas)), np.log(sigmas)))
|
|
else:
|
|
raise ValueError(f"interpolation_type must be one of 'linear' or 'log_linear', got {self.config.interpolation_type}")
|
|
|
|
# 2. Sigma Schedule
|
|
if self.config.use_karras_sigmas:
|
|
sigma_min = self.config.sigma_min if self.config.sigma_min is not None else sigmas[-1]
|
|
sigma_max = self.config.sigma_max if self.config.sigma_max is not None else sigmas[0]
|
|
rho = self.config.rho
|
|
ramp = np.linspace(0, 1, num_inference_steps)
|
|
sigmas = (sigma_max ** (1 / rho) + ramp * (sigma_min ** (1 / rho) - sigma_max ** (1 / rho))) ** rho
|
|
elif self.config.use_exponential_sigmas:
|
|
sigma_min = self.config.sigma_min if self.config.sigma_min is not None else sigmas[-1]
|
|
sigma_max = self.config.sigma_max if self.config.sigma_max is not None else sigmas[0]
|
|
sigmas = np.exp(np.linspace(np.log(sigma_max), np.log(sigma_min), num_inference_steps))
|
|
elif self.config.use_beta_sigmas:
|
|
sigma_min = self.config.sigma_min if self.config.sigma_min is not None else sigmas[-1]
|
|
sigma_max = self.config.sigma_max if self.config.sigma_max is not None else sigmas[0]
|
|
alpha, beta = 0.6, 0.6
|
|
ramp = np.linspace(0, 1, num_inference_steps)
|
|
try:
|
|
import torch.distributions as dist
|
|
|
|
b = dist.Beta(alpha, beta)
|
|
ramp = b.sample((num_inference_steps,)).sort().values.numpy()
|
|
except Exception:
|
|
pass
|
|
sigmas = sigma_max * (1 - ramp) + sigma_min * ramp
|
|
elif self.config.use_flow_sigmas:
|
|
sigmas = np.linspace(1.0, 1 / 1000, num_inference_steps)
|
|
|
|
# 3. Shifting
|
|
if self.config.use_dynamic_shifting and mu is not None:
|
|
sigmas = mu * sigmas / (1 + (mu - 1) * sigmas)
|
|
elif self.config.shift is not None:
|
|
sigmas = self.config.shift * sigmas / (1 + (self.config.shift - 1) * sigmas)
|
|
|
|
# We handle multi-history expansion
|
|
_a_mat, _b_vec, c_vec = self._get_tableau()
|
|
len(c_vec)
|
|
|
|
sigmas_expanded = []
|
|
for i in range(len(sigmas) - 1):
|
|
s_curr = sigmas[i]
|
|
s_next = sigmas[i + 1]
|
|
for c_val in c_vec:
|
|
sigmas_expanded.append(s_curr + c_val * (s_next - s_curr))
|
|
sigmas_expanded.append(0.0)
|
|
|
|
sigmas_interpolated = np.array(sigmas_expanded)
|
|
# Linear remapping for Flow Matching
|
|
timesteps_expanded = sigmas_interpolated * self.config.num_train_timesteps
|
|
|
|
self.sigmas = torch.from_numpy(sigmas_interpolated).to(device=device, dtype=dtype)
|
|
self.timesteps = torch.from_numpy(timesteps_expanded + self.config.steps_offset).to(device=device, dtype=dtype)
|
|
self.init_noise_sigma = self.sigmas.max().item() if self.sigmas.numel() > 0 else 1.0
|
|
|
|
self.model_outputs = []
|
|
self.sample_at_start_of_step = None
|
|
self._step_index = None
|
|
|
|
@property
|
|
def step_index(self):
|
|
"""
|
|
The index counter for the current timestep. It will increase 1 after each scheduler step.
|
|
"""
|
|
return self._step_index
|
|
|
|
def index_for_timestep(self, timestep, schedule_timesteps=None):
|
|
if schedule_timesteps is None:
|
|
schedule_timesteps = self.timesteps
|
|
|
|
if isinstance(schedule_timesteps, torch.Tensor):
|
|
schedule_timesteps = schedule_timesteps.detach().cpu().numpy()
|
|
|
|
if isinstance(timestep, torch.Tensor):
|
|
timestep = timestep.detach().cpu().numpy()
|
|
|
|
return np.abs(schedule_timesteps - timestep).argmin().item()
|
|
|
|
def scale_model_input(self, sample: torch.Tensor, timestep: Union[float, torch.Tensor]) -> torch.Tensor:
|
|
if self._step_index is None:
|
|
self._init_step_index(timestep)
|
|
if self.config.prediction_type == "flow_prediction":
|
|
return sample
|
|
sigma = self.sigmas[self._step_index]
|
|
return sample / ((sigma**2 + 1) ** 0.5)
|
|
|
|
|
|
|
|
def _init_step_index(self, timestep):
|
|
if self._step_index is None:
|
|
if isinstance(timestep, torch.Tensor):
|
|
timestep = timestep.to(self.timesteps.device)
|
|
self._step_index = self.index_for_timestep(timestep)
|
|
|
|
def step(
|
|
self,
|
|
model_output: torch.Tensor,
|
|
timestep: Union[float, torch.Tensor],
|
|
sample: torch.Tensor,
|
|
return_dict: bool = True,
|
|
) -> Union[SchedulerOutput, Tuple]:
|
|
if self._step_index is None:
|
|
self._init_step_index(timestep)
|
|
a_mat, b_vec, c_vec = self._get_tableau()
|
|
num_stages = len(c_vec)
|
|
|
|
stage_index = self._step_index % num_stages
|
|
base_step_index = (self._step_index // num_stages) * num_stages
|
|
|
|
sigma_curr = self.sigmas[base_step_index]
|
|
sigma_next_idx = min(base_step_index + num_stages, len(self.sigmas) - 1)
|
|
sigma_next = self.sigmas[sigma_next_idx]
|
|
|
|
if sigma_next <= 0:
|
|
sigma_t = self.sigmas[self._step_index]
|
|
denoised = sample - sigma_t * model_output if getattr(self.config, "prediction_type", "epsilon") == "epsilon" else model_output
|
|
prev_sample = denoised
|
|
self._step_index += 1
|
|
if not return_dict:
|
|
return (prev_sample,)
|
|
return SchedulerOutput(prev_sample=prev_sample)
|
|
|
|
h = sigma_next - sigma_curr
|
|
sigma_t = self.sigmas[self._step_index]
|
|
|
|
prediction_type = getattr(self.config, "prediction_type", "epsilon")
|
|
if prediction_type == "epsilon":
|
|
denoised = sample - sigma_t * model_output
|
|
elif self.config.prediction_type == "v_prediction":
|
|
alpha_t = 1 / (sigma_t**2 + 1) ** 0.5
|
|
sigma_actual = sigma_t * alpha_t
|
|
denoised = alpha_t * sample - sigma_actual * model_output
|
|
elif prediction_type == "flow_prediction":
|
|
denoised = sample - sigma_t * model_output
|
|
elif prediction_type == "sample":
|
|
denoised = model_output
|
|
else:
|
|
raise ValueError(f"prediction_type error: {getattr(self.config, 'prediction_type', 'epsilon')}")
|
|
|
|
if self.config.clip_sample:
|
|
denoised = denoised.clamp(-self.config.sample_max_value, self.config.sample_max_value)
|
|
|
|
# derivative = (x - x0) / sigma
|
|
derivative = (sample - denoised) / sigma_t if sigma_t > 1e-6 else torch.zeros_like(sample)
|
|
|
|
if self.sample_at_start_of_step is None:
|
|
if stage_index > 0:
|
|
# Mid-step fallback for Img2Img/Inpainting
|
|
sigma_next_t = self.sigmas[self._step_index + 1]
|
|
dt = sigma_next_t - sigma_t
|
|
prev_sample = sample + dt * derivative
|
|
self._step_index += 1
|
|
if not return_dict:
|
|
return (prev_sample,)
|
|
return SchedulerOutput(prev_sample=prev_sample)
|
|
|
|
self.sample_at_start_of_step = sample
|
|
self.model_outputs = [derivative] * stage_index
|
|
|
|
if stage_index == 0:
|
|
self.model_outputs = [derivative]
|
|
self.sample_at_start_of_step = sample
|
|
else:
|
|
self.model_outputs.append(derivative)
|
|
|
|
next_stage_idx = stage_index + 1
|
|
if next_stage_idx < num_stages:
|
|
sum_ak = 0
|
|
for j in range(len(self.model_outputs)):
|
|
sum_ak = sum_ak + a_mat[next_stage_idx][j] * self.model_outputs[j]
|
|
|
|
sigma_next_stage = self.sigmas[self._step_index + 1]
|
|
|
|
# Update x (unnormalized sample)
|
|
prev_sample = self.sample_at_start_of_step + (sigma_next_stage - sigma_curr) * sum_ak
|
|
else:
|
|
sum_bk = 0
|
|
for j in range(len(self.model_outputs)):
|
|
sum_bk = sum_bk + b_vec[j] * self.model_outputs[j]
|
|
|
|
prev_sample = self.sample_at_start_of_step + h * sum_bk
|
|
|
|
self.model_outputs = []
|
|
self.sample_at_start_of_step = None
|
|
|
|
self._step_index += 1
|
|
|
|
if not return_dict:
|
|
return (prev_sample,)
|
|
return SchedulerOutput(prev_sample=prev_sample)
|
|
|
|
def add_noise(
|
|
self,
|
|
original_samples: torch.Tensor,
|
|
noise: torch.Tensor,
|
|
timesteps: torch.Tensor,
|
|
) -> torch.Tensor:
|
|
from .scheduler_utils import add_noise_to_sample
|
|
return add_noise_to_sample(original_samples, noise, self.sigmas, timesteps, self.timesteps)
|
|
|
|
def __len__(self):
|
|
return self.config.num_train_timesteps
|