Add automotive benchmarking setting

[pgmpablo157321]

• Add server_constant_gen parameter to the test_settings
• Add server_constant_gen parameter to the python API
• Let server_constant_gen be able to be loaded from the config files (mlperf.conf and user.conf) using FromConfig
• For the server scenario, the query scheduling distribution from poisson to constant when the parameter server_constant_gen is set to 1, i.e:
  • When server_constant_gen = 0 number of queries in a fixed time follows a poisson distribution (with mean qps) and time between queries follows an exponential distribution (with mean 1/qps).
  • When server_constant_gen = 1 time between queries is constant equal to 1/qps
• Custom Sample Grouping
• In order to run the a benchmark in a grouped setting:
  • Construct a QSL using ConstructGroupedQSL
    • Pass a vector that contains the group sizes in the order they appear in the dataset
  • Start running the test using StartTestWithGroupedQSL
  • Set the variable use_grouped_qsl in the mlperf.conf
  • Run your benchmark in the server scenario

[github-actions[bot]]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅