tensorflow模型冻结

查看模型输入输出节点名称

saved_model_cli show --dir ./xxx --all
如：

冻结模型

freeze_graph --input_saved_model_dir --output_node_names --output_graph

问题

模型冻结时出现的BUG：
tensorflow.python.framework.errors_impl.InternalError: Dst tensor is not initialized.
[[{{node save/RestoreV2/_379}} = _Recvclient_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name=“edge_383_save/RestoreV2”, tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]]
[[{{node save/RestoreV2/_226}} = _SendT=DT_FLOAT, client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name=“edge_231_save/RestoreV2”, _device="/job:localhost/replica:0/task:0/device:CPU:0"]]

经查，据说是超出显存，所以命令前限制GPU解决此问题：CUDA_VISIBLE_DEVICES=’’