metadata

title: ML.ENERGY Leaderboard
emoji: ⚡
python_version: '3.9'
app_file: app.py
sdk: gradio
sdk_version: 3.39.0
pinned: true
tags:
  - energy
  - leaderboard

ML.ENERGY Leaderboard

How much energy do LLMs consume?

This README focuses on explaining how to run the benchmark yourself. The actual leaderboard is here: https://ml.energy/leaderboard.

Colosseum

We instrumented Hugging Face TGI so that it measures and returns GPU energy consumption. Then, our controller server receives user prompts from the Gradio app, selects two models randomly, and streams model responses back with energy consumption.

Setup for benchmarking

Model weights

For models that are directly accessible in Hugging Face Hub, you don't need to do anything.
For other models, convert them to Hugging Face format and put them in /data/leaderboard/weights/lmsys/vicuna-13B, for example. The last two path components (e.g., lmsys/vicuna-13B) are taken as the name of the model.

Docker container

We have our pre-built Docker image published with the tag mlenergy/leaderboard:latest (Dockerfile).

$ docker run -it \
    --name leaderboard0 \
    --gpus '"device=0"' \
    -v /path/to/your/data/dir:/data/leaderboard \
    -v $(pwd):/workspace/leaderboard \
    mlenergy/leaderboard:latest bash

The container internally expects weights to be inside /data/leaderboard/weights (e.g., /data/leaderboard/weights/lmsys/vicuna-7B), and sets the Hugging Face cache directory to /data/leaderboard/hfcache. If needed, the repository should be mounted to /workspace/leaderboard to override the copy of the repository inside the container.

Running the benchmark

We run benchmarks using multiple nodes and GPUs using Pegasus. Take a look at pegasus/ for details.

You can still run benchmarks without Pegasus like this:

$ docker exec leaderboard0 python scripts/benchmark.py --model-path /data/leaderboard/weights/lmsys/vicuna-13B --input-file sharegpt/sg_90k_part1_html_cleaned_lang_first_sampled_sorted.json
$ docker exec leaderboard0 python scripts/benchmark.py --model-path databricks/dolly-v2-12b --input-file sharegpt/sg_90k_part1_html_cleaned_lang_first_sampled_sorted.json

Citation

For the ML.ENERGY Leaderboard:

@misc{ml-energy-leaderboard,
    author={Jae-Won Chung and Jiachen Liu and Zhiyu Wu and Yuxuan Xia and Mosharaf Chowdhury},
    title={{ML.ENERGY} Leaderboard},
    year={2023},
    howpublished={\url{https://ml.energy/leaderboard}},
}

For the Zeus framework used for benchmarking:

@inproceedings{zeus:nsdi23,
    author={Jie You and Jae-Won Chung and Mosharaf Chowdhury},
    booktitle={USENIX NSDI},
    title={Zeus: Understanding and Optimizing {GPU} Energy Consumption of {DNN} Training},
    year={2023},
}