MLPerf v6.1 AMD 8xMI355X_2xEPYC_9575F llama2-70b-99.9 Interactive run_1 harness summary log

Fields

id

sources/bench-mlperf-v6-1-amd-mi355x-llama2-interactive-log

URL

https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/main/closed/AMD/results/8xMI355X_2xEPYC_9575F/llama2-70b-99.9/Interactive/performance/run_1/mlperf_log_summary.txt

Publisher

MLCommons

Kind

benchmark

Accessed

2026-10-04

notes

THE KEY ARTIFACT FOR THE SCENARIO GOTCHA. Result is : VALID, Completed tokens per second 73191.51 - but the log's own header line reads 'Scenario : Server', not 'Scenario : Interactive'. The Interactive result is stored in a directory named Interactive and published in a summary.csv column named Interactive, while the harness log calls it Server and prints 'Performance constraints satisfied: Yes'. Nothing inside the log distinguishes it from a Server run; the ttft<=450ms / tpot<=40ms bounds are compiled into the harness, not recorded in the log. A reader auditing only the raw logs cannot tell an Interactive run from a Server run.

Notes

THE KEY ARTIFACT FOR THE SCENARIO GOTCHA. Result is : VALID, Completed tokens per second 73191.51 - but the log's own header line reads 'Scenario : Server', not 'Scenario : Interactive'. The Interactive result is stored in a directory named Interactive and published in a summary.csv column named Interactive, while the harness log calls it Server and prints 'Performance constraints satisfied: Yes'. Nothing inside the log distinguishes it from a Server run; the ttft<=450ms / tpot<=40ms bounds are compiled into the harness, not recorded in the log. A reader auditing only the raw logs cannot tell an Interactive run from a Server run.

Cross-references

This record references no other records.