How would you know the output node name?
jageshmaharjan opened this issue · comments
jagesh maharjan (马家杰) commented
How would you know the output node name is import/dense_3/Sigmoid
However the model.summary()
is quite not exactly that name.
jagesh maharjan (马家杰) commented
ps: without having to load the tensorboard and checking tf_board graph.
jagesh maharjan (马家杰) commented
Remembered the simplest way (knew this):
tf.saved_model.simple_save(
keras.backend.get_session(), dir_path+'/versionNo',
inputs={'input': model.input},
outputs={t.name: t for t in model.outputs}
)
And, from the terminal you can check input_node and output_node as:
saved_model_cli show --dir Misc/versionNo/ --all
The ouput is:
signature_def['serving_default']:
The given SavedModel SignatureDef contains the following input(s):
inputs['input'] tensor_info:
dtype: DT_FLOAT
shape: (-1, 784)
name: l1_input_input:0
The given SavedModel SignatureDef contains the following output(s):
outputs['last/Softmax:0'] tensor_info:
dtype: DT_FLOAT
shape: (-1, 10)
name: last/Softmax:0
Method name is: tensorflow/serving/predict