fasterbench 0.0.5

A library for benchmarking AI models

fasterbench

Overview

fasterbench is a powerful benchmarking library designed to help AI researchers and engineers evaluate PyTorch models across multiple dimensions:

• Size: Model disk size and parameter count
• Speed: GPU and CPU latency and throughput
• Compute: MACs (multiply-accumulate operations)
• Memory: GPU memory usage
• Energy: Power consumption and carbon emissions

Whether you’re optimizing for deployment, comparing architectures, or researching model efficiency, fasterbench provides the metrics you need to make informed decisions.

Installation

pip install fasterbench

Quick Start

import torch
from torchvision.models import resnet18
from fasterbench import benchmark

# Load your model
model = resnet18()

# Create sample input
dummy_input = torch.randn(1, 3, 224, 224)

# Run comprehensive benchmarks
results = benchmark(model, dummy_input)

Features

• All-in-one benchmarking: Get comprehensive metrics with a single function call
• GPU and CPU performance: Compare inference speed across different hardware
• Environmental impact: Measure carbon footprint with CodeCarbon integration
• Memory profiling: Track peak and average GPU memory usage
• Model comparison: Easily visualize differences between model variants

Example: Comparing Models

from fasterbench import compare_models
from torchvision.models import resnet18, resnet34, resnet50

# Define your models
models = [resnet18(), resnet34(), resnet50()]

# Compare metrics across models
compare_models(models, dls)

Documentation

For more detailed usage examples and API documentation, visit our documentation.

Contributing

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

License

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