QPyTorch: A Low-Precision Arithmetic Simulation Framework

Low-precision training reduces computational cost and produces efficient models. Recent research in developing new low-precision training algorithms often relies on simulation to empirically evaluate the statistical effects of quantization while avoiding the substantial overhead of building specific...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2019 Fifth Workshop on Energy Efficient Machine Learning and Cognitive Computing - NeurIPS Edition (EMC2-NIPS) S. 10 - 13
Hauptverfasser: Zhang, Tianyi, Lin, Zhiqiu, Yang, Guandao, De Sa, Christopher
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.12.2019
Schlagworte:
Computational modeling
Conferences
Energy efficiency
Hardware
low-precision arithmetic
low-precision training
Machine learning
Quantization (signal)
Training
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Low-precision training reduces computational cost and produces efficient models. Recent research in developing new low-precision training algorithms often relies on simulation to empirically evaluate the statistical effects of quantization while avoiding the substantial overhead of building specific hardware. To support this empirical research, we introduce QPyTorch, a low-precision arithmetic simulation framework. Built natively in PyTorch, QPyTorch provides a convenient interface that minimizes the efforts needed to reliably convert existing codes to study low-precision training. QPyTorch is general, and supports a variety of combinations of precisions, number formats, and rounding options. Additionally, it leverages an efficient fused-kernel approach to reduce simulator overhead, which enables simulation of large-scale, realistic problems. QPyTorch is publicly available at https://github.com/Tiiiger/QPyTorch.
DOI:10.1109/EMC2-NIPS53020.2019.00010