提交 b6c5b6f6 编写于 作者: Q qingqing01 提交者: GitHub

Add depth for ResNetC5 in some configs (#2655)

上级 5438754f
......@@ -25,6 +25,7 @@ ResNet:
freeze_at: 2
ResNetC5:
depth: 101
norm_type: affine_channel
RPNHead:
......
......@@ -25,6 +25,7 @@ ResNet:
freeze_at: 2
ResNetC5:
depth: 50
norm_type: affine_channel
RPNHead:
......
......@@ -25,6 +25,7 @@ ResNet:
freeze_at: 2
ResNetC5:
depth: 50
norm_type: affine_channel
RPNHead:
......
......@@ -26,6 +26,7 @@ ResNet:
variant: d
ResNetC5:
depth: 50
norm_type: affine_channel
variant: d
......
......@@ -28,6 +28,7 @@ ResNet:
freeze_at: 2
ResNetC5:
depth: 50
norm_type: affine_channel
RPNHead:
......
......@@ -29,6 +29,7 @@ ResNet:
freeze_at: 2
ResNetC5:
depth: 50
norm_type: affine_channel
RPNHead:
......
......@@ -183,7 +183,8 @@ def load_and_fusebn(exe, prog, path):
[scale_name, bias_name, mean_name, variance_name])
if not bn_in_path:
raise ValueError("The model in path {} has not params of batch norm.")
raise ValueError("There is no params of batch norm in model {}.".format(
path))
# load running mean and running variance on cpu place into global scope.
place = fluid.CPUPlace()
......
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