Lighteval documentation

Installation

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v0.9.2).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Installation

Lighteval can be installed from PyPI or from source. This guide covers all installation options and dependencies.

System Requirements

  • Python: 3.10 or higher
  • PyTorch: 2.0 or higher (but less than 3.0)
  • CUDA: Optional, for GPU acceleration

From PyPI

The simplest way to install Lighteval is from PyPI:

pip install lighteval

This installs the core package with all essential dependencies for basic evaluation tasks.

From Source

Source installation is recommended for developers who want to contribute to Lighteval or need the latest features:

git clone https://github.com/huggingface/lighteval.git
cd lighteval
pip install -e .

Optional Dependencies (Extras)

Lighteval provides several optional dependency groups that you can install based on your needs. Use the format pip install lighteval[<group>] or pip install -e .[<group>] for source installation.

Backend Extras

Extra Description Dependencies
vllm Use VLLM as backend for high-performance inference vllm>=0.10.0, ray, more_itertools
tgi Use Text Generation Inference API text-generation>=0.6.0
litellm Use LiteLLM for unified API access litellm, diskcache
optimum Use Optimum for optimized models optimum==1.12.0
quantization Evaluate quantized models bitsandbytes>=0.41.0, auto-gptq>=0.4.2
adapters Evaluate adapter models (PEFT, Delta) peft==0.3.0
nanotron Evaluate Nanotron models nanotron, tensorboardX

Task and Feature Extras

Extra Description Dependencies
extended_tasks Extended evaluation tasks langdetect, openai>1.87, tiktoken
multilingual Multilingual evaluation support stanza, spacy[ja,ko,th], jieba, pyvi
math Mathematical reasoning tasks latex2sympy2_extended==1.0.6

Storage and Logging Extras

Extra Description Dependencies
s3 Upload results to S3 s3fs
tensorboardX Upload results to TensorBoard tensorboardX
wandb Log results to Weights & Biases wandb
trackio Log results to Trackio trackio

Development Extras

Extra Description Dependencies
quality Code quality tools ruff>=v0.11.0, pre-commit
tests Testing dependencies pytest>=7.4.0, deepdiff
docs Documentation building hf-doc-builder, watchdog
dev All development dependencies Includes accelerate, quality, tests, multilingual, math, extended_tasks, vllm
< > Update on GitHub