What the difference between relay.build_module.create_executor(appear in url) and tvm.contrib.graph_executor.GraphModule(appear in url)?
I have an onnx model encountered error when execute:
intrp = relay.build_module.create_executor("graph", mod, tvm.cpu(0), target)
tmp=intrp.evaluate()
but use tvm.contrib.graph_executor.GraphModule, the model can be ran.