caffe2.python.optimizer.YellowFinOptimizer Class Reference

Inheritance diagram for caffe2.python.optimizer.YellowFinOptimizer:

Public Member Functions
def	__init__ (self, alpha=0.1, mu=0.0, beta=0.999, curv_win_width=20, zero_debias=True, epsilon=0.1 **6, policy='fixed', sparse_dedup_aggregator=None, kwargs)
def	scale_learning_rate (self, scale)
Public Member Functions inherited from caffe2.python.optimizer.Optimizer
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)

Public Attributes
	alpha
	mu
	beta
	curv_win_width
	zero_debias
	epsilon
	policy
	sparse_dedup_aggregator
	init_kwargs

Additional Inherited Members
Static Public Member Functions inherited from caffe2.python.optimizer.Optimizer
def	dedup (net, sparse_dedup_aggregator, grad)

Detailed Description

YellowFin: An automatic tuner for momentum SGD

See https://arxiv.org/abs/1706.03471 for more details. This implementation
has separate learning rate and momentum per each parameter.

Definition at line 1012 of file optimizer.py.

---

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

• caffe2/python/optimizer.py