Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!5966 add bias for Conv2dBnFoldQuant 

Merge pull request !5966 from yuchaojie/quant
This commit is contained in:
mindspore-ci-bot 2020-09-11 12:29:10 +08:00 committed by Gitee
parent 9ca16d3c6c 393a3f5825
commit c021f2379e
1 changed files with 15 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
mindspore/nn/layer/quant.py
View File

@ -419,8 +419,11 @@ class Conv2dBnFoldQuant(Cell):
padding (int): Implicit paddings on both sides of the input. Default: 0. padding (int): Implicit paddings on both sides of the input. Default: 0.
eps (float): Parameters for BatchNormal. Default: 1e-5. eps (float): Parameters for BatchNormal. Default: 1e-5.
momentum (float): Parameters for BatchNormal op. Default: 0.997. momentum (float): Parameters for BatchNormal op. Default: 0.997.
has_bias (bool): Specifies whether the layer uses a bias vector. Default: False.
weight_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the weight_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the
convolution kernel. Default: 'normal'. convolution kernel. Default: 'normal'.
bias_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the
bias vector. Default: 'zeros'.
beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the
beta vector. Default: 'zeros'. beta vector. Default: 'zeros'.
gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the
@ -460,7 +463,9 @@ class Conv2dBnFoldQuant(Cell):
group=1, group=1,
eps=1e-5, eps=1e-5,
momentum=0.997, momentum=0.997,
has_bias=False,
weight_init='normal', weight_init='normal',
bias_init='zeros',
beta_init='zeros', beta_init='zeros',
gamma_init='ones', gamma_init='ones',
mean_init='zeros', mean_init='zeros',
@ -484,6 +489,7 @@ class Conv2dBnFoldQuant(Cell):
self.group = group self.group = group
self.eps = eps self.eps = eps
self.momentum = momentum self.momentum = momentum
self.has_bias = has_bias
self.quant_delay = quant_delay self.quant_delay = quant_delay
self.freeze_bn = freeze_bn self.freeze_bn = freeze_bn
self.fake = fake self.fake = fake
@ -516,6 +522,11 @@ class Conv2dBnFoldQuant(Cell):
weight_shape = [out_channels, in_channels // group, *self.kernel_size] weight_shape = [out_channels, in_channels // group, *self.kernel_size]
channel_axis = 0 channel_axis = 0
self.weight = Parameter(initializer(weight_init, weight_shape), name='weight') self.weight = Parameter(initializer(weight_init, weight_shape), name='weight')
self.bias_add = P.BiasAdd()
if check_bool(has_bias):
self.bias = Parameter(initializer(bias_init, [out_channels]), name='bias')
else:
self.bias = None
# initialize BatchNorm Parameter # initialize BatchNorm Parameter
self.gamma = Parameter(initializer(gamma_init, [out_channels]), name='gamma') self.gamma = Parameter(initializer(gamma_init, [out_channels]), name='gamma')
@ -562,6 +573,8 @@ class Conv2dBnFoldQuant(Cell):
def construct(self, x): def construct(self, x):
out_conv = self.conv(x, self.weight) out_conv = self.conv(x, self.weight)
if self.has_bias:
out_conv = self.bias_add(out_conv, self.bias)
# BN fold1 # BN fold1
batch_mean, batch_std, running_mean, running_std = self.batchnorm_fold(out_conv, batch_mean, batch_std, running_mean, running_std = self.batchnorm_fold(out_conv,
self.moving_mean, self.moving_mean,
@ -572,6 +585,8 @@ class Conv2dBnFoldQuant(Cell):
if self.fake: if self.fake:
weight = self.fake_quant_weight(weight) weight = self.fake_quant_weight(weight)
out = self.conv(x, weight) out = self.conv(x, weight)
if self.has_bias:
out = self.bias_add(out, self.bias)
# BN fold2 # BN fold2
if self.is_gpu: if self.is_gpu:
if self.training: if self.training: