MLPerf Automotive Benchmark Data Download

Quick Start: Copy one of the commands below and run it in your terminal to download the indicated dataset.

Available Downloads

UniAD Benchmark

(click to expand)

UniAD (Tiny) Model

UniAD (Tiny) model for the UniAD MLPerf Automotive Benchmark (~806MB)

            bash <(curl -s https://raw.githubusercontent.com/mlcommons/r2-downloader/refs/heads/main/mlc-r2-downloader.sh) https://automotive.mlcommons-storage.org/metadata/nuscenes_uniad.uri
        

Llama 3.1 8B Benchmark

(click to expand)

Measuring Massive Multitask Language Understanding (MMLU)

A test dataset consisting of multiple choice questions for a variety of subjects to test a LLM's general knowledge. For the automotive benchmark, this tests an LLM building block as a component of an in-vehicle infotainment system. (~271MB)

            bash <(curl -s https://raw.githubusercontent.com/mlcommons/r2-downloader/refs/heads/main/mlc-r2-downloader.sh) -d mmlu/ https://automotive.mlcommons-storage.org/metadata/mmlu.uri
        