Using Concatenation layer in TensorRT 3 RC (Python API)


I was trying to add the concat layer to a network in TRT3 (using the Python API). I am using the following syntax:

concat0 = network.add_concatenation([conv0.get_output(0), conv1.get_output(0)], 2)

And I get the following error:
TypeError: in method ‘NetworkDefinition_add_concatenation’, argument 2 of type ‘nvinfer1::ITensor *const *’

I am not sure if this is a bug or if I am not using the correct syntax as the documentation is still not complete.
Any help will be much appreciated. Thanks!

I also encountered this issue, have you ever solved it?

I submitted a bug report for it and it was accpeted as a bug. It hasn’t been fixed yet.

Same problem, and I worked around it using two auxiliary convs and elementwise add.

BTW, where can we submit BUG reports?

Login to and go to your profile and thats where you submit bugs… yep.