paper-with-me

Papers

Overcoming Oscillations in Quantization-Aware Training

2022-03-21 · Markus Nagel, Marios Fournarakis, Yelysei Bondarenko, Tijmen Blankevoort

When training neural networks with simulated quantization, we observe that quantized weights can, rather unexpectedly, oscillate between two grid-points. The importance of this effect and its impact on quantization-aware training (QAT) are not well-understood or investigated in literature. In this paper, we delve deeper into the phenomenon of weight oscillations and show that it can lead to a significant accuracy degradation due to wrongly estimated batch-normalization statistics during inference and increased noise during training. These effects are particularly pronounced in low-bit ($\leq$ 4-bits) quantization of efficient networks with depth-wise separable layers, such as MobileNets and EfficientNets. In our analysis we investigate several previously proposed QAT algorithms and show that most of these are unable to overcome oscillations. Finally, we propose two novel QAT algorithms to overcome oscillations during training: oscillation dampening and iterative weight freezing. We demonstrate that our algorithms achieve state-of-the-art accuracy for low-bit (3 & 4 bits) weight and activation quantization of efficient architectures, such as MobileNetV2, MobileNetV3, and EfficentNet-lite on ImageNet. Our source code is available at {https://github.com/qualcomm-ai-research/oscillations-qat}.

📄 PDF Abstract BibTeX arXiv:2203.11086

Code (1)

qualcomm-ai-research/oscillations-qat 공식 구현 pytorch

Tasks

Quantization

Methods 이 논문이 사용한 방법론

Pointwise Convolution Pointwise Convolution is a type of convolution that uses a 1x1 kernel: a kernel that iterates through every single point. This…
Depthwise Convolution Depthwise Convolution is a type of convolution where we apply a single convolutional filter for each input channel. In the regular 2D…
Depthwise Separable Convolution While standard convolution performs the channelwise and spatial-wise computation in one step, Depthwise Separable Convolution
Global Average Pooling Global Average Pooling is a pooling operation designed to replace fully connected layers in classical CNNs. The idea is to generate one feature map for each corresponding…
ReLU6 ReLU6 is a modification of the rectified linear unit where we limit the activation to a maximum size of $6$. This is due to increased…
Sigmoid Activation 설명 없음
Batch Normalization 설명 없음
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…

Similar Papers 제목 키워드 기반

FEMBA on the Edge: Physiologically-Aware Pre-Training, Quantization, and Deployment of a Bidirectional Mamba EEG Foundation Model on an Ultra-low Power Microcontroller

2026-03-18 · Anna Tegon, Nicholas Lehmann, Yawei Li, Andrea Cossettini 외 arxiv

Objective: To enable continuous, long-term neuro-monitoring on wearable devices by overcoming the computational bottlenecks of Transformer-based Electroencephalography (EEG) foundation models and the quantization challen…

Singularity-aware Optimization via Randomized Geometric Probing: Towards Stable Non-smooth Optimization

2026-05-28 · Ruoran Xu, Borong She, Xiaobo Jin, Qiufeng Wang arxiv

Deep learning optimization relies heavily on the assumption of smooth loss landscapes, a condition systematically violated by modern architectures due to non-smooth components such as ReLU activations and quantization op…

One-Step Forward and Backtrack: Overcoming Zig-Zagging in Loss-Aware Quantization Training

2024-01-30 · Lianbo Ma, Yuee Zhou, Jianlun Ma, Guo Yu 외

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 …

Quantization

Understanding and Overcoming the Challenges of Efficient Transformer Quantization

2021-09-27 · EMNLP 2021 11 · Yelysei Bondarenko, Markus Nagel, Tijmen Blankevoort

Transformer-based architectures have become the de-facto standard models for a wide range of Natural Language Processing tasks. However, their memory footprint and high latency are prohibitive for efficient deployment an…

Quantization

Overcoming Forgetting Catastrophe in Quantization-Aware Training

2023-01-01 · ICCV 2023 1 · Ting-An Chen, De-Nian Yang, Ming-Syan Chen

Quantization is an effective approach for memory cost reduction by compressing networks to lower bits. However, existing quantization processes learned only from the current data tend to suffer from forgetting catast…

Lifelong learningQuantization