Timezone: »

Training Deep Neural Networks with 8-bit Floating Point Numbers
Naigang Wang · Jungwook Choi · Daniel Brand · Chia-Yu Chen · Kailash Gopalakrishnan

Tue Dec 04 02:00 PM -- 04:00 PM (PST) @ Room 517 AB #108

The state-of-the-art hardware platforms for training deep neural networks are moving from traditional single precision (32-bit) computations towards 16 bits of precision - in large part due to the high energy efficiency and smaller bit storage associated with using reduced-precision representations. However, unlike inference, training with numbers represented with less than 16 bits has been challenging due to the need to maintain fidelity of the gradient computations during back-propagation. Here we demonstrate, for the first time, the successful training of deep neural networks using 8-bit floating point numbers while fully maintaining the accuracy on a spectrum of deep learning models and datasets. In addition to reducing the data and computation precision to 8 bits, we also successfully reduce the arithmetic precision for additions (used in partial product accumulation and weight updates) from 32 bits to 16 bits through the introduction of a number of key ideas including chunk-based accumulation and floating point stochastic rounding. The use of these novel techniques lays the foundation for a new generation of hardware training platforms with the potential for 2-4 times improved throughput over today's systems.

Author Information

Naigang Wang (IBM T. J. Watson Research Center)
Jungwook Choi (IBM Research)
Daniel Brand (IBM Research)
Chia-Yu Chen (IBM research)

my research areas focus on: accelerator architecture compiler design and library development machine learning and neural network VLSI and nano device

Kailash Gopalakrishnan (IBM Research)

More from the Same Authors