def gluon_batchnorm(x,
gamma,
beta,
moving_mean,
moving_var,
momentum=0.9,
axis=1,
epsilon=1e-5,
fix_gamma=False):
"""
Apply native MXNet/Gluon batch normalization on x with given moving_mean, moving_var, beta and gamma.
Parameters
----------
x : keras.backend tensor/variable/symbol
Input tensor/variable/symbol.
gamma : keras.backend tensor/variable/symbol
Tensor by which to scale the input.
beta : keras.backend tensor/variable/symbol
Tensor by which to center the input.
moving_mean : keras.backend tensor/variable/symbol
Moving mean.
moving_var : keras.backend tensor/variable/symbol
Moving variance.
momentum : float, default 0.9
Momentum for the moving average.
axis : int, default 1
Axis along which BatchNorm is applied. Axis usually represent axis of 'channels'. MXNet follows
'channels_first'.
epsilon : float, default 1e-5
Small float added to variance to avoid dividing by zero.
fix_gamma : bool, default False
Fix gamma while training.
Returns
-------
keras.backend tensor/variable/symbol
Resulted tensor/variable/symbol.
"""
if isinstance(x, KerasSymbol):
x = x.symbol
if isinstance(moving_mean, KerasSymbol):
moving_mean = moving_mean.symbol
if isinstance(moving_var, KerasSymbol):
moving_var = moving_var.symbol
if isinstance(beta, KerasSymbol):
beta = beta.symbol
if isinstance(gamma, KerasSymbol):
gamma = gamma.symbol
return KerasSymbol(
mx.sym.BatchNorm(data=x,
gamma=gamma,
beta=beta,
moving_mean=moving_mean,
moving_var=moving_var,
momentum=momentum,
axis=axis,
eps=epsilon,
fix_gamma=fix_gamma))
def gluon_lrn(x, alpha, beta, k, n):
if isinstance(x, KerasSymbol):
x = x.symbol
if isinstance(alpha, KerasSymbol):
alpha = alpha.symbol
if isinstance(beta, KerasSymbol):
beta = beta.symbol
if isinstance(k, KerasSymbol):
k = k.symbol
if isinstance(n, KerasSymbol):
n = n.symbol
return KerasSymbol(
mx.sym.LRN(data=x, alpha=alpha, beta=beta, knorm=k, nsize=n))