Functions — github.com/alisiahkoohi/Langevin-dynamics

↓ 6 callersMethodsample

(self)

examples/gaussian_LD.py:34

↓ 4 callersMethodlr_fn

(t: float, a: float = a, b: float = b, gamma: float = ga

src/langevin_sampling/samplers.py:265

↓ 2 callersMethodnl_pdf

(self, x)

examples/gaussian_LD.py:30

↓ 2 callersMethodproposal_dist

Calculate the proposal distribution for Metropolis-Hastings. Args: idx (int): Index of the current tensor. Retu

src/langevin_sampling/samplers.py:218

↓ 2 callersMethodsample

(self)

examples/gaussian_MALA.py:33

↓ 2 callersMethodstep

Performs a single optimization step. Arguments: closure (callable, optional): A closure that reevaluates the model

src/langevin_sampling/SGLD.py:43

↓ 1 callersMethod__init__

(self, x: torch.Tensor, func: callable, lr: float = 1e-2,

src/langevin_sampling/samplers.py:30

↓ 1 callersMethoddecay_fn

Calculate the learning rate decay function. Args: lr (float): Initial learning rate. lr_final (float): Final

src/langevin_sampling/samplers.py:64

↓ 1 callersMethoddecay_fn

Generate a learning rate decay function. Args: lr (float, optional): Initial learning rate. Default is 1e-2.

src/langevin_sampling/samplers.py:245

↓ 1 callersMethodlr_decay

Update the learning rate of the optimizer based on the current iteration.

src/langevin_sampling/samplers.py:103

↓ 1 callersMethodlr_decay

Decay the learning rate of the optimizer.

src/langevin_sampling/samplers.py:273

↓ 1 callersMethodlr_fn

Calculate the learning rate based on the iteration number. Args: t (float): Current iteration number.

src/langevin_sampling/samplers.py:83

↓ 1 callersMethodsample_prob

Calculate the acceptance probability for Metropolis-Hastings. Returns: torch.Tensor: The acceptance probability.

src/langevin_sampling/samplers.py:235

Method__init__

(self, x: torch.Tensor, func: callable, lr: float = 1e-2,

src/langevin_sampling/samplers.py:144

Method__init__

(self, params, lr=required, momentum=0, da

src/langevin_sampling/SGLD.py:13

Method__init__

Initializes the pSGLD optimizer. Args: params (iterable): Iterable of parameters to optimize. lr (float, opt

src/langevin_sampling/precondSGLD.py:12

Method__init__

(self, mu, cov, device='cuda')

examples/gaussian_LD.py:19

Method__init__

(self, mu, cov, device='cuda')

examples/gaussian_MALA.py:18

Method__setstate__

(self, state)

src/langevin_sampling/SGLD.py:38

Method__setstate__

(self, state)

src/langevin_sampling/precondSGLD.py:50

Methodnl_pdf

(self, x)

examples/gaussian_MALA.py:29

Functionrosenbrock_negative_log

(x)

examples/banana_LD.py:18

Functionrosenbrock_negative_log

(x)

examples/banana_MALA.py:18

Methodsample

Perform a Langevin dynamics step. Returns: tuple: A tuple containing the current parameter values and the loss

src/langevin_sampling/samplers.py:48

Methodsample

Perform a Metropolis-Hastings step to generate a sample. Returns: tuple: A tuple containing the sampled input tensor and

src/langevin_sampling/samplers.py:188

Methodstep

Performs a single optimization step. Args: closure (callable, optional): A closure that reevaluates the model and

src/langevin_sampling/precondSGLD.py:55

Functions26 in github.com/alisiahkoohi/Langevin-dynamics