提交 fbe15e64 编写于 作者: J jiangjiajun

optimize for depthwise conv2d

上级 e1fa0be6
......@@ -503,6 +503,7 @@ class TFOpMapper(OpMapper):
"stride": strides[2:4],
"dilation": dilations[2:4],
"groups": k_size[3] * in_shape[1],
"use_cudnn": False,
"padding": padding
}
node.fluid_code.add_layer("conv2d",
......
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