Toward INT4 Fixed-Point Training via Exploring Quantization Error for Gradients

• URL: http://arxiv.org/abs/2407.12637v1
• Date: Wed, 17 Jul 2024 15:06:12 GMT
• Title: Toward INT4 Fixed-Point Training via Exploring Quantization Error for Gradients
• Authors: Dohyung Kim, Junghyup Lee, Jeimin Jeon, Jaehyeon Moon, Bumsub Ham,
• Abstract summary: We show that lowering the error for large-magnitude gradients boosts the quantization performance significantly. We also introduce an interval update algorithm that adjusts the quantization interval adaptively to maintain a small quantization error for large gradients.
• Score: 24.973203825917906
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Network quantization generally converts full-precision weights and/or activations into low-bit fixed-point values in order to accelerate an inference process. Recent approaches to network quantization further discretize the gradients into low-bit fixed-point values, enabling an efficient training. They typically set a quantization interval using a min-max range of the gradients or adjust the interval such that the quantization error for entire gradients is minimized. In this paper, we analyze the quantization error of gradients for the low-bit fixed-point training, and show that lowering the error for large-magnitude gradients boosts the quantization performance significantly. Based on this, we derive an upper bound of quantization error for the large gradients in terms of the quantization interval, and obtain an optimal condition for the interval minimizing the quantization error for large gradients. We also introduce an interval update algorithm that adjusts the quantization interval adaptively to maintain a small quantization error for large gradients. Experimental results demonstrate the effectiveness of our quantization method for various combinations of network architectures and bit-widths on various tasks, including image classification, object detection, and super-resolution.

Related papers

• One-Step Forward and Backtrack: Overcoming Zig-Zagging in Loss-Aware Quantization Training [12.400950982075948]
  Weight quantization is an effective technique to compress deep neural networks for their deployment on edge devices with limited resources. Traditional loss-aware quantization methods commonly use the quantized gradient to replace the full-precision gradient. This paper proposes a one-step forward and backtrack way for loss-aware quantization to get more accurate and stable gradient direction.
  arXiv  Detail & Related papers  (2024-01-30T05:42:54Z)
• Neural Networks with Quantization Constraints [111.42313650830248]
  We present a constrained learning approach to quantization training. We show that the resulting problem is strongly dual and does away with gradient estimations. We demonstrate that the proposed approach exhibits competitive performance in image classification tasks.
  arXiv  Detail & Related papers  (2022-10-27T17:12:48Z)
• Efficient factored gradient descent algorithm for quantum state tomography [10.100843479138222]
  We present an efficient quantum tomography protocol that combines the state-factored with eigenvalue mapping to address the rank-deficient issue. We also find that our method can accomplish the full-state tomography of random 11-qubit mixed states within one minute.
  arXiv  Detail & Related papers  (2022-07-12T06:45:35Z)
• Cluster-Promoting Quantization with Bit-Drop for Minimizing Network Quantization Loss [61.26793005355441]
  Cluster-Promoting Quantization (CPQ) finds the optimal quantization grids for neural networks. DropBits is a new bit-drop technique that revises the standard dropout regularization to randomly drop bits instead of neurons. We experimentally validate our method on various benchmark datasets and network architectures.
  arXiv  Detail & Related papers  (2021-09-05T15:15:07Z)
• Distribution Adaptive INT8 Quantization for Training CNNs [12.708068468737286]
  In this paper, we propose a novel INT8 quantization training framework for convolutional neural network. Specifically, we adopt Gradient Vectorized Quantization to quantize the gradient, based on the observation that layer-wise gradients contain multiple distributions along the channel dimension. Then, Magnitude-aware Clipping Strategy is introduced by taking the magnitudes of gradients into consideration when minimizing the quantization error.
  arXiv  Detail & Related papers  (2021-02-09T11:58:10Z)
• Adaptive Quantization of Model Updates for Communication-Efficient Federated Learning [75.45968495410047]
  Communication of model updates between client nodes and the central aggregating server is a major bottleneck in federated learning. Gradient quantization is an effective way of reducing the number of bits required to communicate each model update. We propose an adaptive quantization strategy called AdaFL that aims to achieve communication efficiency as well as a low error floor.
  arXiv  Detail & Related papers  (2021-02-08T19:14:21Z)
• Direct Quantization for Training Highly Accurate Low Bit-width Deep Neural Networks [73.29587731448345]
  This paper proposes two novel techniques to train deep convolutional neural networks with low bit-width weights and activations. First, to obtain low bit-width weights, most existing methods obtain the quantized weights by performing quantization on the full-precision network weights. Second, to obtain low bit-width activations, existing works consider all channels equally.
  arXiv  Detail & Related papers  (2020-12-26T15:21:18Z)
• DAQ: Distribution-Aware Quantization for Deep Image Super-Resolution Networks [49.191062785007006]
  Quantizing deep convolutional neural networks for image super-resolution substantially reduces their computational costs. Existing works either suffer from a severe performance drop in ultra-low precision of 4 or lower bit-widths, or require a heavy fine-tuning process to recover the performance. We propose a novel distribution-aware quantization scheme (DAQ) which facilitates accurate training-free quantization in ultra-low precision.
  arXiv  Detail & Related papers  (2020-12-21T10:19:42Z)
• AUSN: Approximately Uniform Quantization by Adaptively Superimposing Non-uniform Distribution for Deep Neural Networks [0.7378164273177589]
  Existing uniform and non-uniform quantization methods exhibit an inherent conflict between the representing range and representing resolution. We propose a novel quantization method to quantize the weight and activation. The key idea is to Approximate the Uniform quantization by Adaptively Superposing multiple Non-uniform quantized values, namely AUSN.
  arXiv  Detail & Related papers  (2020-07-08T05:10:53Z)
• Optimal Gradient Quantization Condition for Communication-Efficient Distributed Training [99.42912552638168]
  Communication of gradients is costly for training deep neural networks with multiple devices in computer vision applications. In this work, we deduce the optimal condition of both the binary and multi-level gradient quantization for textbfANY gradient distribution. Based on the optimal condition, we develop two novel quantization schemes: biased BinGrad and unbiased ORQ for binary and multi-level gradient quantization respectively.
  arXiv  Detail & Related papers  (2020-02-25T18:28:39Z)
• Post-Training Piecewise Linear Quantization for Deep Neural Networks [13.717228230596167]
  Quantization plays an important role in the energy-efficient deployment of deep neural networks on resource-limited devices. We propose a piecewise linear quantization scheme to enable accurate approximation for tensor values that have bell-shaped distributions with long tails. Compared to state-of-the-art post-training quantization methods, our proposed method achieves superior performance on image classification, semantic segmentation, and object detection with minor overhead.
  arXiv  Detail & Related papers  (2020-01-31T23:47:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.