How to track schedule from c++ to python?

1 
          schedule = anchor_impl->second.Schedule(anchor_attrs_, tensor_outs,
                                                  target_);

Here I’m learning the lowering code that create schedule from relay function. When I track to this function, I get lost,

te::Schedule OpImplementation::Schedule(const Attrs& attrs, const Array<te::Tensor>& outs,
                                        const Target& target) {
  return (*this)->fschedule(attrs, outs, target);
}

Seems it tracks into python? How can I find it?

2

I think the current design of lowering on mainline is indeed very much confusing. A better version exists in Relax in the apache/tvm:unity branch, where lowering becomes trivial, and you can take a look at its source code in apache/tvm’s unity branch

3

Thank you! I wonder what’s the status of the relax infra? Does its functionality totally cover that of relay?

And also, is the relax support for dynamic shape available for most models? I didn’t find enough documents for it.