Since we can’t know the cuda kernel args list, does TVM still keep the input and output args continuous, and by topological order?
Since we can’t know the cuda kernel args list, does TVM still keep the input and output args continuous, and by topological order?