run-benchmark

Configuring and launching a benchmark suite with paired-run execution requires applying strict run policies and pre-launch infrastructure checks. You initiate a paired per-suite run by specifying the target suite and parallel slot allocation.

To rerun failed benchmark tasks, you reconfigure the specific suite with desired parameters and initiate a new benchmark execution. Post-run curation steps are then required to ensure the resulting data is valid for analysis.

Yes, pre-launch infrastructure checks are required before executing benchmark runs. These mandatory checks validate the environment to ensure consistency and policy adherence before any benchmark execution begins.

Executing official or gap-fill benchmark batches is constrained by strict run policies, including mandatory paired execution and gap-filling requirements. You must also complete post-run curation steps before valid analysis can occur.

Yes, you can configure parallel slots when launching a benchmark execution. For example, you can launch a paired per-suite run for a specific suite like pytorch with 4 parallel slots to manage execution concurrency.