Compare commits

...

4 Commits

Author SHA1 Message Date
Nathan Lambert
bbd9043be4 add sketch of tests (need more changes) 2022-11-29 17:05:51 -08:00
Nathan Lambert
01b0b868a4 fix copies 2022-10-27 17:13:04 -07:00
Nathan Lambert
f163bccc4e style 2022-10-27 10:59:59 -07:00
Nathan Lambert
864d7b846e init langevin dynamics basic sampler 2022-10-27 10:59:29 -07:00
5 changed files with 318 additions and 0 deletions

View File

@@ -39,6 +39,7 @@ if is_torch_available():
ScoreSdeVePipeline,
)
from .schedulers import (
ALDScheduler,
DDIMScheduler,
DDPMScheduler,
IPNDMScheduler,

View File

@@ -17,6 +17,7 @@ from ..utils import is_flax_available, is_scipy_available, is_torch_available
if is_torch_available():
from .scheduling_ald import ALDScheduler
from .scheduling_ddim import DDIMScheduler
from .scheduling_ddpm import DDPMScheduler
from .scheduling_ipndm import IPNDMScheduler

View File

@@ -0,0 +1,194 @@
# Copyright 2022 UC Berkeley Team and The HuggingFace Team. All rights reserved.
#
# 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.
from dataclasses import dataclass
from typing import Optional, Tuple, Union
import numpy as np
import torch
from ..configuration_utils import ConfigMixin, register_to_config
from ..utils import BaseOutput
from .scheduling_utils import SchedulerMixin
@dataclass
class ALDSchedulerOutput(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.
pred_original_sample (`torch.FloatTensor` 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.FloatTensor
pred_original_sample: Optional[torch.FloatTensor] = None
class ALDScheduler(SchedulerMixin, ConfigMixin):
"""
The Annealed Langevin Dynamics sampler was popularized in the paper on Noise Conditional Score Networks (NCSNs).
For more details, refer to the paper https://arxiv.org/abs/1907.05600
[`~ConfigMixin`] takes care of storing all config attributes that are passed in the scheduler's `__init__`
function, such as `num_train_timesteps`. They can be accessed via `scheduler.config.num_train_timesteps`.
[`~ConfigMixin`] also provides general loading and saving functionality via the [`~ConfigMixin.save_config`] and
[`~ConfigMixin.from_config`] functions.
For more details, see the original paper: https://arxiv.org/abs/2006.11239
Args:
num_train_timesteps (`int`): number of diffusion steps used to train the model.
sigma_min (`float`):
initial noise scale for sigma sequence in sampling procedure. The minimum sigma should mirror the
distribution of the data.
sigma_max (`float`): maximum value used for the range of continuous timesteps passed into the model.
step_lr (`float`): learning rate for stepping through noise.
"""
@register_to_config
def __init__(
self,
num_train_timesteps: int = 100,
sigma_min: float = 0.01,
sigma_max: float = 1.0,
step_lr: float = 0.00002,
):
# standard deviation of the initial noise distribution
self.final_noise_sigma = None
self.step_lr = step_lr
# setable values
self.num_inference_steps = None
self.timesteps = torch.from_numpy(np.arange(0, num_train_timesteps)[::-1].copy())
self.set_sigmas(num_train_timesteps, sigma_min, sigma_max)
def scale_model_input(self, sample: torch.FloatTensor, timestep: Optional[int] = None) -> torch.FloatTensor:
"""
Ensures interchangeability with schedulers that need to scale the denoising model input depending on the
current timestep.
Args:
sample (`torch.FloatTensor`): input sample
timestep (`int`, optional): current timestep
Returns:
`torch.FloatTensor`: scaled input sample
"""
return sample
def set_timesteps(self, num_inference_steps: int, device: Union[str, torch.device] = None):
"""
Sets the continuous timesteps used for the diffusion chain. Supporting function to be run before inference.
Args:
num_inference_steps (`int`):
the number of diffusion steps used when generating samples with a pre-trained model.
"""
num_inference_steps = min(self.config.num_train_timesteps, num_inference_steps)
self.num_inference_steps = num_inference_steps
timesteps = np.arange(
0, self.config.num_train_timesteps, self.config.num_train_timesteps // self.num_inference_steps
)[::-1].copy()
self.timesteps = torch.from_numpy(timesteps).to(device)
def set_sigmas(self, num_inference_steps: int, sigma_min: float = None, sigma_max: float = None):
"""
Sets the noise scales used for the diffusion chain. Supporting function to be run before inference.
The sigmas control the weight of the `drift` and `diffusion` components of sample update.
Args:
num_inference_steps (`int`):
the number of diffusion steps used when generating samples with a pre-trained model.
sigma_min (`float`, optional):
initial noise scale value (overrides value given at Scheduler instantiation).
sigma_max (`float`, optional): final noise scale value (overrides value given at Scheduler instantiation).
"""
sigma_min = sigma_min if sigma_min is not None else self.config.sigma_min
sigma_max = sigma_max if sigma_max is not None else self.config.sigma_max
if self.timesteps is None:
self.set_timesteps(num_inference_steps)
self.sigmas = torch.tensor(
np.exp(np.linspace(np.log(sigma_min), np.log(sigma_max), num_inference_steps)),
dtype=torch.float32,
)
self.final_noise_sigma = self.sigmas[-1]
def step(
self,
model_output: torch.FloatTensor,
timestep: int,
sample: torch.FloatTensor,
generator=None,
return_dict: bool = True,
) -> Union[ALDSchedulerOutput, Tuple]:
"""
Predict the sample at the previous timestep by reversing the SDE. Core function to propagate the diffusion
process from the learned model outputs (most often the predicted noise).
Args:
model_output (`torch.FloatTensor`): direct output from learned diffusion model.
timestep (`int`): current discrete timestep in the diffusion chain.
sample (`torch.FloatTensor`):
current instance of sample being created by diffusion process.
generator: random number generator.
return_dict (`bool`): option for returning tuple rather than ALDSchedulerOutput class
Returns:
[`~schedulers.scheduling_utils.ALDSchedulerOutput`] or `tuple`:
[`~schedulers.scheduling_utils.ALDSchedulerOutput`] if `return_dict` is True, otherwise a `tuple`. When
returning a tuple, the first element is the sample tensor.
"""
t = timestep
# 1. get sigma
sigma = self.sigmas[t]
# 2. compute step_size
step_size = self.step_lr * (sigma / self.final_noise_sigma) ** 2
# 3. create new output
pred_prev_sample = sample + step_size * model_output
# 4. Add noise except last step
variance = 0
if t > 0:
noise = torch.randn(
model_output.size(), dtype=model_output.dtype, layout=model_output.layout, generator=generator
).to(model_output.device)
variance = noise * torch.sqrt(step_size * 2)
pred_prev_sample = pred_prev_sample + variance
if not return_dict:
return (pred_prev_sample,)
return ALDSchedulerOutput(prev_sample=pred_prev_sample)
def __len__(self):
return self.config.num_train_timesteps

View File

@@ -242,6 +242,21 @@ class ScoreSdeVePipeline(metaclass=DummyObject):
requires_backends(cls, ["torch"])
class ALDScheduler(metaclass=DummyObject):
_backends = ["torch"]
def __init__(self, *args, **kwargs):
requires_backends(self, ["torch"])
@classmethod
def from_config(cls, *args, **kwargs):
requires_backends(cls, ["torch"])
@classmethod
def from_pretrained(cls, *args, **kwargs):
requires_backends(cls, ["torch"])
class DDIMScheduler(metaclass=DummyObject):
_backends = ["torch"]

View File

@@ -20,6 +20,7 @@ import numpy as np
import torch
from diffusers import (
ALDScheduler,
DDIMScheduler,
DDPMScheduler,
IPNDMScheduler,
@@ -875,6 +876,112 @@ class ScoreSdeVeSchedulerTest(unittest.TestCase):
self.assertEqual(output_0.shape, sample.shape)
self.assertEqual(output_0.shape, output_1.shape)
class ALDSchedulerTest(unittest.TestCase):
# TODO adapt with class SchedulerCommonTest (scheduler needs Numpy Integration), similar to SDE VE
scheduler_classes = (ALDScheduler,)
forward_default_kwargs = ()
@property
def dummy_sample(self):
batch_size = 4
num_channels = 3
height = 8
width = 8
sample = torch.rand((batch_size, num_channels, height, width))
return sample
@property
def dummy_sample_deter(self):
batch_size = 4
num_channels = 3
height = 8
width = 8
num_elems = batch_size * num_channels * height * width
sample = torch.arange(num_elems)
sample = sample.reshape(num_channels, height, width, batch_size)
sample = sample / num_elems
sample = sample.permute(3, 0, 1, 2)
return sample
def dummy_model(self):
def model(sample, t, *args):
return sample * t / (t + 1)
return model
def get_scheduler_config(self, **kwargs):
config = {
"num_train_timesteps": 100,
"sigma_min": 0.01,
"sigma_max": 1.0,
"step_lr": 0.00002,
}
config.update(**kwargs)
return config
def check_over_configs(self, time_step=0, **config):
kwargs = dict(self.forward_default_kwargs)
for scheduler_class in self.scheduler_classes:
sample = self.dummy_sample
residual = 0.1 * sample
scheduler_config = self.get_scheduler_config(**config)
scheduler = scheduler_class(**scheduler_config)
with tempfile.TemporaryDirectory() as tmpdirname:
scheduler.save_config(tmpdirname)
new_scheduler = scheduler_class.from_config(tmpdirname)
output = scheduler.step(
residual, time_step, sample, generator=torch.manual_seed(0), **kwargs
).prev_sample
new_output = new_scheduler.step(
residual, time_step, sample, generator=torch.manual_seed(0), **kwargs
).prev_sample
assert torch.sum(torch.abs(output - new_output)) < 1e-5, "Scheduler outputs are not identical"
def check_over_forward(self, time_step=0, **forward_kwargs):
kwargs = dict(self.forward_default_kwargs)
kwargs.update(forward_kwargs)
for scheduler_class in self.scheduler_classes:
sample = self.dummy_sample
residual = 0.1 * sample
scheduler_config = self.get_scheduler_config()
scheduler = scheduler_class(**scheduler_config)
with tempfile.TemporaryDirectory() as tmpdirname:
scheduler.save_config(tmpdirname)
new_scheduler = scheduler_class.from_config(tmpdirname)
output = scheduler.step(
residual, time_step, sample, generator=torch.manual_seed(0), **kwargs
).prev_sample
new_output = new_scheduler.step(
residual, time_step, sample, generator=torch.manual_seed(0), **kwargs
).prev_sample
assert torch.sum(torch.abs(output - new_output)) < 1e-5, "Scheduler outputs are not identical"
def test_timesteps(self):
for timesteps in [10, 100, 1000]:
self.check_over_configs(num_train_timesteps=timesteps)
def test_sigmas(self):
for sigma_min, sigma_max in zip([0.0001, 0.001, 0.01], [1, 1, 1]):
self.check_over_configs(sigma_min=sigma_min, sigma_max=sigma_max)
def test_time_indices(self):
for t in [0.1, 0.5, 0.75]:
self.check_over_forward(time_step=t)
class LMSDiscreteSchedulerTest(SchedulerCommonTest):
scheduler_classes = (LMSDiscreteScheduler,)