TVM on Jetson TX2 (GPU)


I am trying to run resnet18 on TX2 (GPU). On CPU it is working fine but on GPU, I am getting a seg fault.

Below is the host and target assignment. I suspect it may not be correct.

target ="nvidia/jetson-tx2", host="llvm -mtriple=aarch64-linux-gnu")
dev = tvm.cuda(0)

with tvm.transform.PassContext(opt_level=3):
    lib =, target=target, params=params)

m = graph_executor.GraphModule(lib["default"](dev))

The last line is causing the Segmentation fault Is there a different way, I should be providing the target and host?

I have tried target ="cuda", host="llvm") and target ="cuda -tx2", host="llvm")

Thank you.