pytorch_pfn_extras.training.triggers.OnceTrigger#
- class pytorch_pfn_extras.training.triggers.OnceTrigger(call_on_resume=False)#
Bases:
Trigger
Trigger based on the starting point of the iteration.
This trigger accepts only once at starting point of the iteration. There are two ways to specify the starting point: only starting point in whole iteration or called again when training resumed.
- Parameters:
call_on_resume (bool) – Whether the extension is called again or not when restored from a snapshot. It is set to
False
by default.
- finished#
Flag that indicates whether or not this trigger will
- Type:
bool
- fire in the future. This flag is used to determine if the extension
- should be initialized after resume.
Methods
__init__
([call_on_resume])load_state_dict
(to_load)may_fire
(iteration, epoch_length)Flags if the trigger may fire at the current iteration
Attributes
- __call__(manager)#
Call self as a function.
- Parameters:
manager (ExtensionsManagerProtocol) –
- Return type:
bool
- __init__(call_on_resume=False)#
- Parameters:
call_on_resume (bool) –
- Return type:
None
- property finished: bool#
- load_state_dict(to_load)#
- Parameters:
to_load (Dict[str, Any]) –
- Return type:
None
- may_fire(iteration, epoch_length)#
Flags if the trigger may fire at the current iteration
This must not alter the trigger state
- Parameters:
iteration (int) –
epoch_length (int) –
- Return type:
bool
- state_dict()#
- Return type:
Dict[str, Any]