MXNet Operator Performance Benchmarks

Motivation

How to use

Prerequisites

Usecase 1 - Run benchmarks for all the operators

Usecase 2 - Run benchmarks for all the operators in a specific category

Usecase 3 - Run benchmarks for specific operator

Usecase 4 - Run benchmarks for group of operators with same input

Usecase 5 - Profile internal operators locally

Usecase 6 - Compare performance for chosen operator from both NDArray library and its Numpy/Numpyextension counterpart

Changes

Old Code

New Code

Result

Example

How does it work under the hood?

Use python timer

TODO