Hugging Face's logo

Spaces:
ml-energy
/
leaderboard
Running

App Files Community
3
leaderboard / README.md
Jae-Won Chung
README title formatting
f0fe800
preview code
|
raw
history blame
1.02 kB

ML.ENERGY Leaderboard

How much energy do LLMs consume?

Devs

Current setup in ampere02:

  1. Find model weights in /data/leaderboard/weights/, e.g. subdirectory llama and vicuna.
  2. Let's share the Huggingface Transformer cache:
export TRANSFORMERS_CACHE=/data/leaderboard/hfcache

Run benchmarks like this:

$ docker build -t leaderboard:latest .
$ docker run -it --name jw-leaderboard --gpus all --cap-add SYS_ADMIN -v /data/leaderboard:/data/leaderboard -v $HOME/workspace/leaderboard:/workspace/leaderboard leaderboard:latest bash

# cd leaderboard
# python scripts/benchmark.py --model-path /data/leaderboard/weights/lmsys/vicuna-7B --input-file /data/leaderboard/sharegpt/sg_90k_part1_html_cleaned_lang_first_sampled.json
# python scripts/benchmark.py --model-path databricks/dolly-v2-12b --input-file /data/leaderboard/sharegpt/sg_90k_part1_html_cleaned_lang_first_sampled.json