Since, all the benchmarks are using very similar kinds of inference interface, and bench.sh files are almost all the same. So, it would better to have a full refactoring of the code base with a better file-structure with re-usable modules.
This should not only help us to benchmark, but also might help us to build an unified inference tool in later versions.
Since, all the benchmarks are using very similar kinds of inference interface, and bench.sh files are almost all the same. So, it would better to have a full refactoring of the code base with a better file-structure with re-usable modules.
This should not only help us to benchmark, but also might help us to build an unified inference tool in later versions.