Lighteval

🤗 Lighteval is your all-in-one toolkit for evaluating Large Language Models (LLMs) across multiple backends with ease. Dive deep into your model’s performance by saving and exploring detailed, sample-by-sample results to debug and see how your models stack up.

Key Features

🚀 Multi-Backend Support

Evaluate your models using the most popular and efficient inference backends:

• transformers: Evaluate models on CPU or one or more GPUs using 🤗 Accelerate
• nanotron: Evaluate models in distributed settings using ⚡️ Nanotron
• vllm: Evaluate models on one or more GPUs using 🚀 VLLM
• custom: Evaluate custom models (can be anything)
• sglang: Evaluate models using SGLang as backend
• inference-endpoint: Evaluate models using Hugging Face’s Inference Endpoints API
• tgi: Evaluate models using 🔗 Text Generation Inference running locally
• litellm: Evaluate models on any compatible API using LiteLLM
• inference-providers: Evaluate models using HuggingFace’s inference providers as backend**: Distributed training and evaluation

📊 Comprehensive Evaluation

• Extensive Task Library: 1000s pre-built evaluation tasks
• Custom Task Creation: Build your own evaluation tasks
• Flexible Metrics: Support for custom metrics and scoring
• Detailed Analysis: Sample-by-sample results for deep insights

🔧 Easy Customization

Customization at your fingertips: create new tasks, metrics or model tailored to your needs, or browse all our existing tasks and metrics.

☁️ Seamless Integration

Seamlessly experiment, benchmark, and store your results on the Hugging Face Hub, S3, or locally.

Quick Start

Installation

pip install lighteval

Basic Usage

# Evaluate a model using Transformers backend
lighteval accelerate \
    "model_name=openai-community/gpt2" \
    "leaderboard|truthfulqa:mc|0"

Save Results

# Save locally
lighteval accelerate \
    "model_name=openai-community/gpt2" \
    "leaderboard|truthfulqa:mc|0" \
    --output-dir ./results

# Push to Hugging Face Hub
lighteval accelerate \
    "model_name=openai-community/gpt2" \
    "leaderboard|truthfulqa:mc|0" \
    --push-to-hub \
    --results-org your-username