caffe2.python.optimizer.Optimizer Class Reference

Inheritance diagram for caffe2.python.optimizer.Optimizer:

Public Member Functions
def	__init__ (self)
def	__call__ (self, net, param_init_net, param, grad=None)
def	get_cpu_blob_name (self, base_str, node_name='')
def	get_gpu_blob_name (self, base_str, gpu_id, node_name)
def	make_unique_blob_name (self, base_str)
def	build_lr (self, net, param_init_net, base_learning_rate, learning_rate_blob=None, policy="fixed", iter_val=0, kwargs)
def	add_lr_multiplier (self, lr_multiplier)
def	get_auxiliary_parameters (self)
def	scale_learning_rate (self, args, kwargs)
def	create_lars_inputs (self, param_init_net, weight_decay, trust, lr_max)

Static Public Member Functions
def	dedup (net, sparse_dedup_aggregator, grad)

Detailed Description

Definition at line 30 of file optimizer.py.

Member Function Documentation

def caffe2.python.optimizer.Optimizer.add_lr_multiplier	(		self,
			lr_multiplier
	)

Set the global learning rate multiplier. If a multiplier already
existed, this will overwrite the existing multiplier. The multiplier is
used for all future calls to _run(), unless it is overwritten.

Definition at line 147 of file optimizer.py.

def caffe2.python.optimizer.Optimizer.get_auxiliary_parameters

(

self

)

Returns a list of auxiliary parameters.

Returns:
    aux_params: A namedtuple, AuxParams.

    aux_params.local stores a list of blobs. Each blob is a local
    auxiliary parameter. A local auxiliary parameter is a parameter in
    parallel to a learning rate parameter. Take adagrad as an example,
    the local auxiliary parameter is the squared sum parameter, because
    every learning rate has a squared sum associated with it.

    aux_params.shared also stores a list of blobs. Each blob is a shared
    auxiliary parameter. A shared auxiliary parameter is a parameter
    that is shared across all the learning rate parameters. Take adam as
    an example, the iteration parameter is a shared parameter, because
    all the learning rates share the same iteration parameter.

Definition at line 181 of file optimizer.py.

def caffe2.python.optimizer.Optimizer.make_unique_blob_name	(		self,
			base_str
	)

Returns a blob name that will be unique to the current device
and optimizer instance.

Definition at line 75 of file optimizer.py.

---

The documentation for this class was generated from the following file:

• caffe2/python/optimizer.py