ONNX and tensorRT: ERROR: Network must have at least one output

[SunilJB]

I have the same error. I make a very simple model which has only one conv3d layer.

class Model(torch.nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.conv = torch.nn.Conv3d(in_channels=1, out_channels=16, kernel_size=3, stride=1, padding=1)

    def forward(self, x):
        return self.conv(x)

Then I generate an onnx model without any error.

def generate_onnx():
    model = Model().cuda()
    dummy_input = torch.randn(1, 1, 112, 112, 112, device='cuda:0', dtype=torch.float)
    torch.onnx.export(model, dummy_input, 'easy_model.onnx', verbose=True, input_names=['input'], output_names=['output'])

After that, I run the following code just same as the sample code in onnx_resnet50.py in introductory_parser_samples.

# The Onnx path is used for Onnx models.
def build_engine_onnx(model_file):
    with trt.Builder(TRT_LOGGER) as builder, builder.create_network() as network, trt.OnnxParser(network, TRT_LOGGER) as parser:
        builder.max_workspace_size = common.GiB(1)
        # Load the Onnx model and parse it in order to populate the TensorRT network.
        with open(model_file, 'rb') as model:
            parser.parse(model.read())
        return builder.build_cuda_engine(network)


if __name__ == '__main__':
    model_path = './easy_model.onnx'
    build_engine_onnx(model_path)

I get [TensorRT] ERROR: Network must have at least one output. Please help. I use pytorch 1.3 and TensorRT 6.0.1.

This specific issue is arising because the ONNX Parser isn’t currently compatible with the ONNX models exported from Pytorch 1.3 - If you downgrade to Pytorch 1.2, this issue should go away.