_ParamScheduler

class mmengine.optim._ParamScheduler(optimizer, param_name, begin=0, end=1000000000, last_step=-1, by_epoch=True, verbose=False)

Base class for parameter schedulers.

It should be inherited by all schedulers that schedule parameters in the optimizer’s param_groups. All subclasses should overwrite the _get_value() according to their own schedule strategy. The implementation is motivated by https://github.com/pytorch/pytorch/blob/master/torch/optim/lr_scheduler.py.

Parameters:

• optimizer (BaseOptimWrapper or Optimizer) – Wrapped optimizer.

• param_name (str) – Name of the parameter to be adjusted, such as lr, momentum.

• begin (int) – Step at which to start updating the parameters. Defaults to 0.

• end (int) – Step at which to stop updating the parameters. Defaults to INF.

• last_step (int) – The index of last step. Used for resuming without state dict. Default value -1 means the step function is never be called before. Defaults to -1.

• by_epoch (bool) – Whether the scheduled parameters are updated by epochs. Defaults to True.

• verbose (bool) – Whether to print the value for each update. Defaults to False.

get_last_value()

Return the last computed value by current scheduler.

Returns:

A list of the last computed value of the optimizer’s param_group.

Return type:

list

load_state_dict(state_dict)

Loads the schedulers state.

Parameters:

state_dict (dict) – scheduler state. Should be an object returned from a call to state_dict().

print_value(is_verbose, group, value)

Display the current parameter value.

Parameters:

• is_verbose (bool) – Whether to print the value.

• group (int) – The index of the current param_group.

• value (float) – The parameter value.

state_dict()

Returns the state of the scheduler as a dict.

It contains an entry for every variable in self.__dict__ which is not the optimizer.

Returns:

scheduler state.

Return type:

dict

step()

Adjusts the parameter value of each parameter group based on the specified schedule.