First, thanks for providing this project - it helped me a lot with using the c_api of Tensorflow in my project. One problem I encountered happened when using load_graph.cpp to import a model containing any time-dependent layer (RNN, GRU, LSTM). The program crashes at TF_SessionRun()with the follwing contents of the status:
Expected input[1] == 'TensorArrayV2_1/element_shape:output:0' to be a control input.
In {{node TensorArrayV2Stack/TensorListStack}}
[[{{node test_model/lstm/PartitionedCall}}]]
[[{{node StatefulPartitionedCall}}]]
[[{{node StatefulPartitionedCall}}]]
Do you have any idea what that is supposed to mean? I cannot understand what is missing here since input[1] is supposed to be 0. If you or anyone else has ever encountered this or know anything about that - I'm so happy about any piece of advice since im stuck with this for way too long now.
Hello!
First, thanks for providing this project - it helped me a lot with using the c_api of Tensorflow in my project. One problem I encountered happened when using load_graph.cpp to import a model containing any time-dependent layer (RNN, GRU, LSTM). The program crashes at
TF_SessionRun()
with the follwing contents of the status:Do you have any idea what that is supposed to mean? I cannot understand what is missing here since input[1] is supposed to be 0. If you or anyone else has ever encountered this or know anything about that - I'm so happy about any piece of advice since im stuck with this for way too long now.
Thanks very much for your help, for a longer describtion of what I'm trying to do I have posted this also on stackoverflow: https://stackoverflow.com/questions/64754471/tensorflow-c-api-1-13-2-import-lstm-pb-expected-input1-to-be-a-control-i