This phase will see the implementation of features to automate or speed-up typical user routines. In particular, we anticipate supporting the following use cases:

• running the benchmark on custom lightgbm builds and/or branches
• analyzing the benchmark results in a systematic script for crunching numbers and compile a readable report
• a nice report format for comparing pairs of lightgbm implementations (ex: master vs branch), or multiple implementations against one.
• systematic testing of every script in a consistent unit test methodology
• identifying ways to standardize code to avoid duplication

Milestone: an end-to-end benchmark run producing report automatically, comparing pairs of variants of lightgbm (ex: stable vs custom build).