Publication

Accuracy Boosters: Epoch-Driven Mixed-Mantissa Block Floating-Point for DNN Training

2023

Abstract

The unprecedented growth in DNN model complexity, size and the amount of training data have led to a commensurate increase in demand for computing and a search for minimal encoding. Recent research advocates Hybrid Block Floating-Point (HBFP) as a technique that minimizes silicon provisioning in accelerators by converting the majority of arithmetic operations in training to 8-bit fixed-point. In this paper, we perform a full-scale exploration of the HBFP design space including minimal mantissa encoding, varying block sizes, and mixed mantissa bit-width across layers and epochs. We propose Accuracy Boosters, an epoch-driven mixedmantissa HBFP that uses 6-bit mantissa only in the last epoch and converts 99.7% of all arithmetic operations in training to 4-bit mantissas. Accuracy Boosters enable reducing silicon provisioning for an HBFP training accelerator by 16.98× as compared to FP32, while preserving or outperforming FP32 accuracy.

Official source

https://infoscience.epfl.ch/entities/publication/0812c290-c9e9-438a-a163-f3d290f544d3

About this result

This page is automatically generated and may contain information that is not correct, complete, up-to-date, or relevant to your search query. The same applies to every other page on this website. Please make sure to verify the information with EPFL's official sources.

Graph Chat

Chat with Graph Search

Ask any question about EPFL courses, lectures, exercises, research, news, etc. or try the example questions below.

DISCLAIMER: The Graph Chatbot is not programmed to provide explicit or categorical answers to your questions. Rather, it transforms your questions into API requests that are distributed across the various IT services officially administered by EPFL. Its purpose is solely to collect and recommend relevant references to content that you can explore to help you answer your questions.

2023

Abstract

Official source

https://infoscience.epfl.ch/entities/publication/0812c290-c9e9-438a-a163-f3d290f544d3

About this result

Related concepts (32)

Half-precision floating-point format

In computing, half precision (sometimes called FP16 or float16) is a binary floating-point computer number format that occupies 16 bits (two bytes in modern computers) in computer memory. It is intended for storage of floating-point values in applications where higher precision is not essential, in particular and neural networks. Almost all modern uses follow the IEEE 754-2008 standard, where the 16-bit base-2 format is referred to as binary16, and the exponent uses 5 bits.

Single-precision floating-point format

Single-precision floating-point format (sometimes called FP32 or float32) is a computer number format, usually occupying 32 bits in computer memory; it represents a wide dynamic range of numeric values by using a floating radix point. A floating-point variable can represent a wider range of numbers than a fixed-point variable of the same bit width at the cost of precision. A signed 32-bit integer variable has a maximum value of 231 − 1 = 2,147,483,647, whereas an IEEE 754 32-bit base-2 floating-point variable has a maximum value of (2 − 2−23) × 2127 ≈ 3.

Fixed-point arithmetic

In computing, fixed-point is a method of representing fractional (non-integer) numbers by storing a fixed number of digits of their fractional part. Dollar amounts, for example, are often stored with exactly two fractional digits, representing the cents (1/100 of dollar). More generally, the term may refer to representing fractional values as integer multiples of some fixed small unit, e.g. a fractional amount of hours as an integer multiple of ten-minute intervals.

Related publications (1)

Accuracy Boosters: Epoch-Driven Mixed-Mantissa Block Floating-Point for DNN Training

Chat with Graph Search

Miniboxing: An Encoding for Specialization