optim package

Subpackages

• optim.scheduler package
  • Submodules
  • optim.scheduler.base_scheduler module
    • BaseLRScheduler
      • BaseLRScheduler.__init__()
      • BaseLRScheduler.add_arguments()
      • BaseLRScheduler.get_lr()
      • BaseLRScheduler.update_lr()
      • BaseLRScheduler.retrieve_lr()
      • BaseLRScheduler.extra_repr()
  • optim.scheduler.cosine module
    • CosineScheduler
      • CosineScheduler.__init__()
      • CosineScheduler.add_arguments()
      • CosineScheduler.get_lr()
  • optim.scheduler.cyclic module
    • CyclicLRScheduler
      • CyclicLRScheduler.__init__()
      • CyclicLRScheduler.add_arguments()
      • CyclicLRScheduler.get_lr()
  • optim.scheduler.fixed module
    • FixedLRScheduler
      • FixedLRScheduler.__init__()
      • FixedLRScheduler.add_arguments()
      • FixedLRScheduler.get_lr()
  • optim.scheduler.multi_step module
    • MultiStepLRScheduler
      • MultiStepLRScheduler.__init__()
      • MultiStepLRScheduler.add_arguments()
      • MultiStepLRScheduler.get_lr()
  • optim.scheduler.polynomial module
    • PolynomialScheduler
      • PolynomialScheduler.__init__()
      • PolynomialScheduler.add_arguments()
      • PolynomialScheduler.get_lr()
  • Module contents
    • build_scheduler()
    • general_lr_sch_args()
    • arguments_scheduler()

Submodules

optim.adam module

class optim.adam.AdamOptimizer(opts: Namespace, model_params: Iterable[Tensor | Dict], *args, **kwargs)

Bases: BaseOptim, Adam

Adam optimizer

Parameters:

• opts – Command-line arguments

• model_params – Model parameters

__init__(opts: Namespace, model_params: Iterable[Tensor | Dict], *args, **kwargs) → None

classmethod add_arguments(parser: ArgumentParser) → ArgumentParser

Add arguments for ADAM optimizer

optim.adamw module

class optim.adamw.AdamWOptimizer(opts: Namespace, model_params: Iterable[Tensor | Dict], *args, **kwargs)

Bases: BaseOptim, AdamW

AdamW optimizer

Parameters:

• opts – Command-line arguments

• model_params – Model parameters

__init__(opts: Namespace, model_params: Iterable[Tensor | Dict], *args, **kwargs) → None

classmethod add_arguments(parser: ArgumentParser) → ArgumentParser

Add arguments for AdamW optimizer

optim.base_optim module

class optim.base_optim.BaseOptim(opts: Namespace, *args, **kwargs)

Bases: object

Base class for optimizer

Parameters:

opts – Command-line arguments

__init__(opts: Namespace, *args, **kwargs) → None

classmethod add_arguments(parser: ArgumentParser) → ArgumentParser

Add optimizer arguments

optim.sgd module

class optim.sgd.SGDOptimizer(opts: Namespace, model_params: Iterable[Tensor | Dict], *args, **kwargs)

Bases: BaseOptim, SGD

SGD optimizer

Parameters:

• opts – Command-line arguments

• model_params – Model parameters

__init__(opts: Namespace, model_params: Iterable[Tensor | Dict], *args, **kwargs) → None

classmethod add_arguments(parser: ArgumentParser) → ArgumentParser

Add optimizer arguments

Module contents

optim.check_trainable_parameters(model: Module, model_params: List) → None

Helper function to check if any model parameters w/ gradients are not part of model_params

optim.remove_param_name_key(model_params: List) → None

Helper function to remove param_names key from model_params

optim.build_optimizer(model: Module, opts, *args, **kwargs) → BaseOptim

Helper function to build an optimizer

Parameters:

• model – A model

• opts – command-line arguments

Returns:

An instance of BaseOptim

optim.arguments_optimizer(parser: ArgumentParser) → ArgumentParser