pytorch_pfn_extras.training.ExtensionsManager¶
- class pytorch_pfn_extras.training.ExtensionsManager(models, optimizers, max_epochs, *, iters_per_epoch, extensions=None, out_dir='result', stop_trigger=None, writer=None, transform_model=<function ExtensionsManager.<lambda>>, enable_profile=False)¶
Manages the extensions and the current status.
- Parameters
models (dict or torch.nn.Module) – Map of string to Module or an actual Module
optimizers (dict or torch.Optimizer) – Map of string to Optimizer or an actual Optimizer.
max_epochs (int) – Number of epochs in the whole training loop. Ignored if stop_trigger is passed as a kwarg.
iters_per_epoch (int) – Number of iterations in one epoch.
extensions (list or None) – List of Extentions to be used.
out_dir (str) – Output directory (default:
result
).stop_trigger (trigger object, optional) – to determine wether training has concluded. The default is an interval trigger set to max_epochs
writer (writing.Writer object) – Writer that can be used by extensions to write data to custom filesystems.
enable_profile (bool) – Flag to enable/disable profiling of iterations. Default is False.
transform_model (Callable[[str, torch.nn.modules.module.Module], torch.nn.modules.module.Module]) –
- Return type
None
- __init__(models, optimizers, max_epochs, *, iters_per_epoch, extensions=None, out_dir='result', stop_trigger=None, writer=None, transform_model=<function ExtensionsManager.<lambda>>, enable_profile=False)¶
- Parameters
models (Union[torch.nn.modules.module.Module, Dict[str, torch.nn.modules.module.Module]]) –
optimizers (Union[torch.optim.optimizer.Optimizer, Dict[str, torch.optim.optimizer.Optimizer]]) –
max_epochs (int) –
iters_per_epoch (int) –
extensions (Optional[Sequence[extension_module.ExtensionLike]]) –
out_dir (str) –
stop_trigger (trigger_module.TriggerLike) –
writer (Optional[pytorch_pfn_extras.writing._writer_base.Writer]) –
transform_model (Callable[[str, torch.nn.modules.module.Module], torch.nn.modules.module.Module]) –
enable_profile (bool) –
- Return type
None
Methods
__init__
(models, optimizers, max_epochs, *, …)complete_iteration
(*[, observation, …])Context manager to complete deferred iterations.
extend
(extension[, name, trigger, priority, …])Registers an extension to the manager.
get_extension
(name)Returns the extension of a given name.
load_state_dict
(to_load)needs_model_state
([iteration])needs_state_this_iteration
()run_extensions
(*[, completed, only_iterations])run_iteration
(*[, step_optimizers])Context manager to run an iteration.
start_extensions
()state_dict
()Attributes
elapsed_time
epoch
epoch_detail
is_before_training
iteration
models
optimizers
out
raw_models
stop_trigger
updater