未验证 提交 93229f1d 编写于 作者: P parap1uie-s 提交者: GitHub

Update resnet_acnet.py (#4297)

Bias attr of square conv should be "False" rather than None during training mode.
上级 8f2e0d59
...@@ -172,7 +172,7 @@ class ResNetACNet(object): ...@@ -172,7 +172,7 @@ class ResNetACNet(object):
act=act if self.deploy else None, act=act if self.deploy else None,
param_attr=ParamAttr(name=name + "_acsquare_weights"), param_attr=ParamAttr(name=name + "_acsquare_weights"),
bias_attr=ParamAttr(name=name + "_acsquare_bias") bias_attr=ParamAttr(name=name + "_acsquare_bias")
if self.deploy else None, if self.deploy else False,
name=name + '.acsquare.conv2d.output.1') name=name + '.acsquare.conv2d.output.1')
if self.deploy: if self.deploy:
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册