I am new to TVM so apology for silly questions. So my current goal is to load a pre-trained model, assign some of the layers (functions?) to one device (say, cpu) and the rest to another device (say, cuda). I am wondering how to achieve that?
Note that I see support of running a model on heterogeneous devices such as #1773, but don’t know how to modify an existing model and send it tvm.build.
Thanks and Regards.