Now can the model run on nvidia without tune?

I haven’t touched TVM in a while. In the previous version of tvm0.9, the error “uses too much shared data” may occur if the model is not compiled through scheduling

1 Like