evallite

EvalLite 🚀

An efficient, zero-cost LLM evaluation framework combining the simplicity of DeepEval with the power of free Hugging Face models through AILite.

🌟 Key Features

• Zero-Cost Evaluation: Leverage free Hugging Face models for LLM evaluation
• Simple Integration: Drop-in replacement for DeepEval's evaluation capabilities
• Extensive Model Support: Access to leading open-source models including:
  • Meta Llama 3.1 70B Instruct
  • Qwen 2.5 72B Instruct
  • Mistral Nemo Instruct
  • Phi-3.5 Mini Instruct
  • And more!
• Comprehensive Metrics: Full compatibility with DeepEval's evaluation metrics
• Async Support: Built-in asynchronous evaluation capabilities

📥 Installation

pip install evallite

🚀 Quick Start

Here's a simple example to get you started with EvalLite:

from evallite import (
    assert_test,
    EvalLiteModel,
    LLMTestCase,
    evaluate,
    AnswerRelevancyMetric
)

# Initialize metric with a specific model
answer_relevancy_metric = AnswerRelevancyMetric(
    threshold=0.7,
    model=EvalLiteModel(model="microsoft/Phi-3.5-mini-instruct")
)

# Create a test case
test_case = LLMTestCase(
    input="What if these shoes don't fit?",
    actual_output="We offer a 30-day full refund at no extra costs.",
    retrieval_context=["All customers are eligible for a 30 day full refund at no extra costs."]
)

# Run evaluation
evaluate([test_case], [answer_relevancy_metric])

🔧 Available Models

EvalLite supports several powerful open-source models:

from evallite import EvalLiteModel

# Available model options
models = [
    'meta-llama/Meta-Llama-3.1-70B-Instruct',
    'CohereForAI/c4ai-command-r-plus-08-2024',
    'Qwen/Qwen2.5-72B-Instruct',
    'nvidia/Llama-3.1-Nemotron-70B-Instruct-HF',
    'meta-llama/Llama-3.2-11B-Vision-Instruct',
    'NousResearch/Hermes-3-Llama-3.1-8B',
    'mistralai/Mistral-Nemo-Instruct-2407',
    'microsoft/Phi-3.5-mini-instruct'
]

# Initialize with specific model
evaluator = EvalLiteModel(model='microsoft/Phi-3.5-mini-instruct')

📊 Advanced Usage

Custom Schema Support

EvalLite supports custom response schemas using Pydantic models:

from pydantic import BaseModel
from typing import List

class Statements(BaseModel):
    statements: List[str]

# Use with schema
result = evaluator.generate(
    prompt="List three facts about climate change",
    schema=Statements
)

Async Evaluation

async def evaluate_async():
    response = await evaluator.a_generate(
        prompt="What is the capital of France?",
        schema=Statements
    )
    return response

Batch Evaluation

from evallite import EvaluationDataset

# Create multiple test cases
test_cases = [
    LLMTestCase(
        input="Question 1",
        actual_output="Answer 1",
        retrieval_context=["Context 1"]
    ),
    LLMTestCase(
        input="Question 2",
        actual_output="Answer 2",
        retrieval_context=["Context 2"]
    )
]

# Create dataset
dataset = EvaluationDataset(test_cases=test_cases)

# Evaluate all at once
evaluate(dataset, [answer_relevancy_metric])

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📄 License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

🙏 Acknowledgments

• DeepEval for the evaluation framework
• AILite for providing free model access
• The open-source community for making powerful language models accessible