研究人员开发了一种名为高效量化前调优（ETBQ）的新方法，以提高深度神经网络低比特训练后量化（PTQ）的精度。该技术包括一个预处理调优阶段，在PTQ过程之前优化全精度模型，使其对量化误差不那么敏感。ETBQ不需要训练假量化模型，因此计算效率很高。在ImageNet和Cityscapes等各种数据集上的实验表明，ETBQ在不同任务中显著提高了低比特PTQ的性能。 AI
影响 提高了在资源受限设备上部署深度神经网络的效率和准确性。
排序理由 该集群包含一篇详细介绍优化深度神经网络新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2607.11359v1 Announce Type: new Abstract: Post-training quantization (PTQ) compresses deep neural networks for deployment under limited memory and computational budgets. However, low-bit (i.e., 2-bit or 4-bit) PTQ often suffers from substantial performance degradation. Most…
Post-training quantization (PTQ) compresses deep neural networks for deployment under limited memory and computational budgets. However, low-bit (i.e., 2-bit or 4-bit) PTQ often suffers from substantial performance degradation. Most existing PTQ methods operate on an unconstraine…