Pychop: Emulating Low-Precision Arithmetic in Numerical Methods and Neural Networks

2025-04-10

AI Summary

To address the challenge of efficiently and controllably simulating low-precision arithmetic in numerical computing and deep learning, this paper introduces Pychop—the first open-source, unified low-precision simulation framework for scientific computing and deep learning. Pychop supports customizable IEEE-compliant floating-point formats, multiple rounding modes, and fine-grained precision configuration. It implements numerically accurate Python-based simulation alongside CUDA-accelerated tensor operations, enabling seamless integration with PyTorch and JAX and GPU-accelerated training and inference. Its key contribution lies in establishing a unified simulation paradigm across frameworks and precision levels, providing a standardized experimental platform for mixed-precision algorithm design and novel hardware validation. Empirical evaluation on image classification and object detection tasks demonstrates substantial reductions in memory footprint and energy consumption, with Top-1 accuracy degradation under 1%. The framework is publicly released and has gained broad adoption in the research community.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Large Vision Models

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Motivated by the growing demand for low-precision arithmetic in computational science, we exploit lower-precision emulation in Python -- widely regarded as the dominant programming language for numerical analysis and machine learning. Low-precision training has revolutionized deep learning by enabling more efficient computation and reduced memory and energy consumption while maintaining model fidelity. To better enable numerical experimentation with and exploration of low precision computation, we developed the Pychop library, which supports customizable floating-point formats and a comprehensive set of rounding modes in Python, allowing users to benefit from fast, low-precision emulation in numerous applications. Pychop also introduces interfaces for both PyTorch and JAX, enabling efficient low-precision emulation on GPUs for neural network training and inference with unparalleled flexibility. In this paper, we offer a comprehensive exposition of the design, implementation, validation, and practical application of Pychop, establishing it as a foundational tool for advancing efficient mixed-precision algorithms. Furthermore, we present empirical results on low-precision emulation for image classification and object detection using published datasets, illustrating the sensitivity of the use of low precision and offering valuable insights into its impact. Pychop enables in-depth investigations into the effects of numerical precision, facilitates the development of novel hardware accelerators, and integrates seamlessly into existing deep learning workflows. Software and experimental code are publicly available at https://github.com/inEXASCALE/pychop.

Problem

• Emulating low-precision arithmetic for efficient computation in Python
• Supporting customizable floating-point formats and rounding modes
• Enabling low-precision neural network training and inference on GPUs

Innovation

• Emulates low-precision arithmetic in Python
• Supports customizable floating-point formats
• Integrates with PyTorch and JAX for GPU emulation