Skip to content

P
PaddleOCR

weixin_41840029 / PaddleOCR
与 Fork 源项目一致

Fork自 PaddlePaddle / PaddleOCR

通知 1
Star 1
Fork 0
P
PaddleOCR
提交 33d96880 编写于 作者: W WenmuZhou
浏览文件
操作

更新NumpyArrayInitializer为Assign

上级 65d3dfc7
隐藏空白更改
内联 并排
Showing with 2 addition and 4 deletion
ppocr/modeling/transform/tps.py
浏览文件 @ 33d96880
...@@ -102,13 +102,11 @@ class LocalizationNetwork(nn.Layer): ...@@ -102,13 +102,11 @@ class LocalizationNetwork(nn.Layer):
name = "loc_fc2" name = "loc_fc2"
param_attr = ParamAttr( param_attr = ParamAttr(
learning_rate=loc_lr, learning_rate=loc_lr,
initializer=paddle.fluid.initializer.NumpyArrayInitializer( initializer=nn.initializer.Assign(np.zeros([fc_dim, F * 2])),
np.zeros([fc_dim, F * 2])),
name=name + "_w") name=name + "_w")
bias_attr = ParamAttr( bias_attr = ParamAttr(
learning_rate=loc_lr, learning_rate=loc_lr,
initializer=paddle.fluid.initializer.NumpyArrayInitializer( initializer=nn.initializer.Assign(initial_bias),
initial_bias),
name=name + "_b") name=name + "_b")
self.fc2 = nn.Linear( self.fc2 = nn.Linear(
fc_dim, fc_dim,
......
Markdown is supported
0% 或 .
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 或