all. I want generate a freezed model from one LSTM model (https://github.com/roatienza/Deep-Learning-Experiments/tree/master/Experiments/Tensorflow/RNN). In my option, I should freeze the last prediction node and use "bazel-bin/tensorflow/python/tools/freeze_graph --input_binary=true --input_graph=model_20170913/model.pb --input_checkpoint=model_20170913/model.ckpt --output_graph=model_20170913/frozen_graph.pb --output_node_names=ArgMax_52"(ArgMax_52 is last default node name). However, I got one notice "Converted 0 variables to const ops." (freeze command's result). Now, I have no idea about which node_name should be as output_node_name?
所有。我想从一个LSTM模型生成一个冻结模型（https://github.com/roatienza/Deep-Learning-Experiments/tree/master/Experiments/Tensorflow/RNN）。在我的选项中，我应该冻结最后一个预测节点并使用“bazel-bin / tensorflow / python / tools / freeze_graph --input_binary = true --input_graph = model_20170913 / model.pb --input_checkpoint = model_20170913 / model.ckpt - output_graph = model_20170913 / frozen_graph.pb --output_node_names = ArgMax_52“（ArgMax_52是最后一个默认节点名称）。但是，我收到一条通知“将0变量转换为const操作”。 （冻结命令的结果）。现在，我不知道哪个node_name应该是output_node_name？
1 个解决方案
#1
0
As mentioned above, "lstm_prediction" is output_node_name. And Tensorboard help me a lot to understand the graph.
如上所述，“lstm_prediction”是output_node_name。 Tensorboard帮助我理解图表。
#1
0
As mentioned above, "lstm_prediction" is output_node_name. And Tensorboard help me a lot to understand the graph.
如上所述，“lstm_prediction”是output_node_name。 Tensorboard帮助我理解图表。