def _resource_apply_dense(self, grad, var, apply_state=None):
var_device, var_dtype = var.device, var.dtype.base_dtype
coefficients = ((apply_state or {}).get((var_device, var_dtype))
or self._fallback_apply_state(var_device, var_dtype))
accum = self.get_slot(var, 'accumulator')
linear = self.get_slot(var, 'linear')
if self._l2_shrinkage_regularization_strength <= 0.0:
return gen_training_ops.ResourceApplyFtrl(
var=var.handle,
accum=accum.handle,
linear=linear.handle,
grad=grad,
lr=coefficients['lr_t'],
l1=coefficients['l1_regularization_strength'],
l2=coefficients['l2_regularization_strength'],
lr_power=coefficients['learning_rate_power'],
use_locking=self._use_locking)
else:
return gen_training_ops.ResourceApplyFtrlV2(
var=var.handle,
accum=accum.handle,
linear=linear.handle,
grad=grad,
lr=coefficients['lr_t'],
l1=coefficients['l1_regularization_strength'],
l2=coefficients['l2_regularization_strength'],
l2_shrinkage=coefficients[
'l2_shrinkage_regularization_strength'],
lr_power=coefficients['learning_rate_power'],
use_locking=self._use_locking)
def _resource_apply_dense(self, grad, var, apply_state=None):
var_device, var_dtype = var.device, var.dtype.base_dtype
coefficients = ((apply_state or {}).get((var_device, var_dtype))
or self._fallback_apply_state(var_device, var_dtype))
# Adjust L2 regularization strength to include beta to avoid the underlying
# TensorFlow ops needing to include it.
adjusted_l2_regularization_strength = (
coefficients['l2_regularization_strength'] + coefficients['beta'] /
(2. * coefficients['lr_t']))
accum = self.get_slot(var, 'accumulator')
linear = self.get_slot(var, 'linear')
if self._l2_shrinkage_regularization_strength <= 0.0:
return gen_training_ops.ResourceApplyFtrl(
var=var.handle,
accum=accum.handle,
linear=linear.handle,
grad=grad,
lr=coefficients['lr_t'],
l1=coefficients['l1_regularization_strength'],
l2=adjusted_l2_regularization_strength,
lr_power=coefficients['learning_rate_power'],
use_locking=self._use_locking)
else:
return gen_training_ops.ResourceApplyFtrlV2(
var=var.handle,
accum=accum.handle,
linear=linear.handle,
grad=grad,
lr=coefficients['lr_t'],
l1=coefficients['l1_regularization_strength'],
l2=adjusted_l2_regularization_strength,
l2_shrinkage=coefficients[
'l2_shrinkage_regularization_strength'],
lr_power=coefficients['learning_rate_power'],
use_locking=self._use_locking)