Hi, all! Hope you are all doing all.
I have a question regarding the type conversion.
I’m trying to convert relay.Var to te.tensor.Tensor in the Python side.
My first idea was to extract shape and type info from relay.Var and create the te.placeholder accordingly.
However, it seems like this approach does not work since I’m losing some necessary attributes. (e.g., operator tag like softmax.op.tag in tvm/python/tvm/topi/cuda/softmax.py)
What would be the best practice in this case?
Any advice or idea would be greatly helpful! Thank you.