Hi all;
I am reading the source codes and try to understand the link between Relay layer and Tensor Expression Layer.
For Tensor Expression(te), I know we can use operators in topi. If I want to change some operators .e.g. extend conv2d_int8 with Tensor Core as backend instead of current dp4a. Then I only need to change the codes in conv22_int8 operators.
However, I did find explicit relation between replay operator with topi operator. For instance, I can find nn.conv2d by printing replay module (tvm.ir.module.IRModule). But how can I know which conv2d in topi will the nn.conv2d use?
Thank you very much!