Hello, using torch.utils.tensorboard and tb-nightly I am having some issues adding a graph to the summary. The model I am using is similar to the rgbd ENet implementation (ENetDepth at https://github.com/montrealrobotics/ENetDepth/blob/master/models/enet.py).
I am getting the error:
UserWarning: ONNX export failed on ATen operator max_unpool2d because torch.onnx.symbolic.max_unpool2d does not exist
.format(op_name, op_name))
The model uses maxunpool layers in the UpsamplingBottlenecks:
self.main_unpool1 = nn.MaxUnpool2d(kernel_size=2)
and forwarded:
# Main branch shortcut
main = self.main_conv1(x)
main = self.main_unpool1(main, max_indices)
# Extension branch
ext = self.ext_conv1(x)
ext = self.ext_conv2(ext)
ext = self.ext_conv3(ext)
ext = self.ext_regul(ext)
# Add main and extension branches
out = main + ext
return self.out_prelu(out)
I was wondering if anyone had any tips on a solution to my problem, or a quick workaround?
Thanks