Created by: dianhuasici

rnn = StaticRNN(name=scope)
layers.Print(input_embedding, message=scope + "/embedding", summarize=0)
embedding = layers.transpose(input_embedding, perm=[1, 0, 2], name=scope)
layers.Print(embedding, message=scope + "/embedding_transpose", summarize=0)

我调用了这段代码两次，输出结果如下：

1548057563	encode_fw/embedding	Tensor[embedding_0.tmp_0]
	shape: [128,30,300,]
	dtype: f
	data: 
1548057563	encode_fw/embedding_transpose	Tensor[encode_fw.tmp_0]
	shape: [30,128,300,]
	dtype: f
	data: 
1548057564	encode_bw/embedding	Tensor[reverse_1.tmp_0]
	shape: [128,30,300,]
	dtype: f
	data: 
1548057564	encode_bw/embedding_transpose	Tensor[encode_fw.tmp_0]
	shape: [30,128,300,]
	dtype: f
	data: 

可以看到第一次我输入进的scope为encode_fw,这时前两个输出的tensor的name都是对的，但是我第二次调用，transpose之后输出的name还是第一次的名字