Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

do not regularize beta and bias #11953

Closed
wants to merge 1 commit into from
Closed

do not regularize beta and bias #11953

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions python/mxnet/gluon/nn/basic_layers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -208,7 +208,7 @@ def __init__(self, units, activation=None, use_bias=True, flatten=True,
if use_bias:
self.bias = self.params.get('bias', shape=(units,),
init=bias_initializer, dtype=dtype,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)
else:
self.bias = None
if activation is not None:
Expand Down Expand Up @@ -334,7 +334,7 @@ def __init__(self, axis=1, momentum=0.9, epsilon=1e-5, center=True, scale=True,
differentiable=scale)
self.beta = self.params.get('beta', grad_req='write' if center else 'null',
shape=(in_channels,), init=beta_initializer,
allow_deferred_init=True,
wd_mult=0.0, allow_deferred_init=True,
differentiable=center)
self.running_mean = self.params.get('running_mean', grad_req='null',
shape=(in_channels,),
Expand Down Expand Up @@ -509,7 +509,7 @@ def __init__(self, axis=1, epsilon=1e-5, center=True, scale=False,
allow_deferred_init=True)
self.beta = self.params.get('beta', grad_req='write' if center else 'null',
shape=(in_channels,), init=beta_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)

def hybrid_forward(self, F, x, gamma, beta):
if self._axis == 1:
Expand Down Expand Up @@ -597,7 +597,7 @@ def __init__(self, axis=-1, epsilon=1e-5, center=True, scale=True,
allow_deferred_init=True)
self.beta = self.params.get('beta', grad_req='write' if center else 'null',
shape=(in_channels,), init=beta_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)

def hybrid_forward(self, F, data, gamma, beta):
norm_data = F.LayerNorm(data, gamma=gamma, beta=beta, axis=self._axis, eps=self._epsilon)
Expand Down
2 changes: 1 addition & 1 deletion python/mxnet/gluon/nn/conv_layers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,7 +118,7 @@ def __init__(self, channels, kernel_size, strides, padding, dilation,
allow_deferred_init=True)
if use_bias:
self.bias = self.params.get('bias', shape=wshapes[2],
init=bias_initializer,
init=bias_initializer, wd_mult=0.0,
allow_deferred_init=True)
else:
self.bias = None
Expand Down
12 changes: 6 additions & 6 deletions python/mxnet/gluon/rnn/rnn_cell.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -369,10 +369,10 @@ def __init__(self, hidden_size, activation='tanh',
allow_deferred_init=True)
self.i2h_bias = self.params.get('i2h_bias', shape=(hidden_size,),
init=i2h_bias_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)
self.h2h_bias = self.params.get('h2h_bias', shape=(hidden_size,),
init=h2h_bias_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)

def state_info(self, batch_size=0):
return [{'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}]
Expand Down Expand Up @@ -482,10 +482,10 @@ def __init__(self, hidden_size,
allow_deferred_init=True)
self.i2h_bias = self.params.get('i2h_bias', shape=(4*hidden_size,),
init=i2h_bias_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)
self.h2h_bias = self.params.get('h2h_bias', shape=(4*hidden_size,),
init=h2h_bias_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)
self._activation = activation
self._recurrent_activation = recurrent_activation

Expand Down Expand Up @@ -597,10 +597,10 @@ def __init__(self, hidden_size,
allow_deferred_init=True)
self.i2h_bias = self.params.get('i2h_bias', shape=(3*hidden_size,),
init=i2h_bias_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)
self.h2h_bias = self.params.get('h2h_bias', shape=(3*hidden_size,),
init=h2h_bias_initializer,
allow_deferred_init=True)
wd_mult=0.0, allow_deferred_init=True)

def state_info(self, batch_size=0):
return [{'shape': (batch_size, self._hidden_size), '__layout__': 'NC'}]
Expand Down
4 changes: 2 additions & 2 deletions python/mxnet/gluon/rnn/rnn_layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,11 +71,11 @@ def __init__(self, hidden_size, num_layers, layout,
self.i2h_bias.append(
self.params.get('%s%d_i2h_bias'%(j, i), shape=(ng*nh,),
init=i2h_bias_initializer,
allow_deferred_init=True))
wd_mult=0.0, allow_deferred_init=True))
self.h2h_bias.append(
self.params.get('%s%d_h2h_bias'%(j, i), shape=(ng*nh,),
init=h2h_bias_initializer,
allow_deferred_init=True))
wd_mult=0.0, allow_deferred_init=True))
ni = nh * self._dir

self._unfused = self._unfuse()
Expand Down