A Study of BFLOAT16 for Deep Learning Training

TL;DR

This paper empirically demonstrates that BFLOAT16, a half-precision floating-point format with the same range as FP32, can be effectively used for deep learning training across various domains without hyper-parameter tuning.

Contribution

It provides the first comprehensive empirical analysis of BFLOAT16 for deep learning training, showing its efficacy and implementation details across multiple frameworks.

Findings

BFLOAT16 achieves the same SOTA results as FP32 in training.

Training with BFLOAT16 requires no hyper-parameter tuning.

BFLOAT16 simplifies conversion and maintains training stability.

Abstract

This paper presents the first comprehensive empirical study demonstrating the efficacy of the Brain Floating Point (BFLOAT16) half-precision format for Deep Learning training across image classification, speech recognition, language modeling, generative networks and industrial recommendation systems. BFLOAT16 is attractive for Deep Learning training for two reasons: the range of values it can represent is the same as that of IEEE 754 floating-point format (FP32) and conversion to/from FP32 is simple. Maintaining the same range as FP32 is important to ensure that no hyper-parameter tuning is required for convergence; e.g., IEEE 754 compliant half-precision floating point (FP16) requires hyper-parameter tuning. In this paper, we discuss the flow of tensors and various key operations in mixed precision training, and delve into details of operations, such as the rounding modes for converting FP32 tensors to BFLOAT16. We have implemented a method to emulate BFLOAT16 operations in Tensorflow, Caffe2, IntelCaffe, and Neon for our experiments. Our results show that deep learning training using BFLOAT16 tensors achieves the same state-of-the-art (SOTA) results across domains as FP32 tensors in the same number of iterations and with no changes to hyper-parameters.