I have been trying to compile a network with transposed 2D convolution layer but no luck so far. I have tried both tf.layers.conv2d_transpose and tf.nn.conv2d_transpose APIs. My code trains well and I can create a meta file but when I try to compile via
mvNCCompile inference.meta -in input -on output -o inference.graph
I get this error:
tensorflow.python.framework.errors_impl.InvalidArgumentError: You must feed a value for placeholder tensor 'input' with dtype float and shape [?,784]
[[node input (defined at /usr/local/bin/ncsdk/Controllers/TensorFlowParser.py:213) = Placeholder[dtype=DT_FLOAT, shape=[?,784], _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]
My OS is Ubuntu 16.04
Tensorflow-GPU Version: 1.9.0
I can share my files I required. I would really appreciate if you could guide me what is the problem.