Training with tvm backward ops on Mxnet

[xqdan]

Hi,

Suppose I have tvm backward ops for a model, how can I train this model on Mxnet with tvm ops?
what’s the whole work like?

Thanks,