TL;DR
This paper introduces HMQ, a hardware-friendly mixed precision quantization block that uses a Gumbel-Softmax estimator to efficiently search for optimal quantization schemes suitable for edge devices, achieving competitive results on CIFAR10 and ImageNet.
Contribution
HMQ is a novel quantization block that enforces uniform, power-of-two thresholds and employs a smooth estimator for efficient search, advancing hardware-compatible CNN quantization.
Findings
Achieves competitive accuracy on CIFAR10 and ImageNet.
Successfully quantizes four different architectures with added restrictions.
Demonstrates state-of-the-art results despite constraints.
Abstract
Recent work in network quantization produced state-of-the-art results using mixed precision quantization. An imperative requirement for many efficient edge device hardware implementations is that their quantizers are uniform and with power-of-two thresholds. In this work, we introduce the Hardware Friendly Mixed Precision Quantization Block (HMQ) in order to meet this requirement. The HMQ is a mixed precision quantization block that repurposes the Gumbel-Softmax estimator into a smooth estimator of a pair of quantization parameters, namely, bit-width and threshold. HMQs use this to search over a finite space of quantization schemes. Empirically, we apply HMQs to quantize classification models trained on CIFAR10 and ImageNet. For ImageNet, we quantize four different architectures and show that, in spite of the added restrictions to our quantization scheme, we achieve competitive and, in…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
MethodsRAdam · Gumbel Softmax
