8. Benchmark grading

Unlike previous years, there will NOT be a benchmarking interview for this year’s competition.

The overall benchmarking score for a team will depend on the output (i.e. performance) obtained for each benchmark and the team’s adherence to the rules set by the SCC committee. The benchmarks (HPL, HPL-MXP, and MLPerf Inference LLAMA2-70B) will carry (30%, 35%, 35%) weights towards the final benchmarking score.