参考tensorflow2.0 C++加载python训练保存的pb模型
经过模型训练及保存,我们得到“OptimalModelDataSet2”
文件夹,模型的保存方法(.h5
或.pb
文件),参考【Visual Studio Code】c/c++部署tensorflow训练的模型
其中“OptimalModelDataSet2”文件夹保存着训练好的模型数据"saved_model.pb"
要在OptimalModelDataSet2所在的文件夹下执行:
$ saved_model_cli show --dir ./OptimalModelDataSet2
会得到:
The given SavedModel contains the following tag-sets:
'serve'
使用'serve',继续执行:
$ saved_model_cli show --dir ./OptimalModelDataSet2/ --tag_set serve
会得到:
The given SavedModel MetaGraphDef contains SignatureDefs with the following keys:
SignatureDef key: "__saved_model_init_op"
SignatureDef key: "serving_default"
使用上述信息,继续:
$ saved_model_cli show --dir ./OptimalModelDataSet2/ --tag_set serve --signature_def serving_default
得到:
The given SavedModel SignatureDef contains the following input(s):inputs['input_1'] tensor_info:dtype: DT_FLOATshape: (-1, 24)name: serving_default_input_1:0
The given SavedModel SignatureDef contains the following output(s):outputs['dense_30'] tensor_info:dtype: DT_FLOATshape: (-1, 1)name: StatefulPartitionedCall:0
Method name is: tensorflow/serving/predict
这里输入和输出的name,要留好,c中调用会用到。