d12e7259c2
* Add explicit copy stage for batch_matmul(x, x) case * do copy in relay strategy to avoid dup * add copy to dense op and schedules * black * add batch_matmul test * add dense test * fix cuda int8 dense test * remove need_copy flag * do not use tag to decide if tensors are same * rename to copy_if_identical and add comment * black * one more fix missed * add length check on input tensors * one more length check * fix variable name