让大模型按需运行，省钱又高效。
FlexRank: Nested Low-Rank Knowledge Decomposition for Adaptive Model Deployment
- 从预训练模型中提取重要性分级的低秩子模块
- 支持不同算力预算下灵活部署，性能随成本平滑提升
- 一次训练，多场景部署，适合资源受限环境
深度神经网络（包括大语言模型和视觉变换器）规模持续扩大，导致从零训练成本过高，部署也愈发昂贵。这些模型通常作为固定计算量的‘计算巨兽’使用，难以适应不同成本预算下的部署需求。我们提出，可从预训练模型中提取按重要性排序的嵌套组件，并根据可用算力选择性激活。为此，我们设计了FlexRank方法，通过基于重要性的低秩权重分解与嵌套整合，提取出能力逐级增强的子模型。该方法实现‘一次训练，处处部署’，在不重新训练的前提下，为不同预算提供性能与成本之间的平稳权衡，推动大模型的实际应用部署。
原文摘要 · Abstract (English)
The growing scale of deep neural networks, encompassing large language models (LLMs) and vision transformers (ViTs), has made training from scratch prohibitively expensive and deployment increasingly costly. These models are often used as computational monoliths with fixed cost, hindering adaptive deployment across different cost budgets. We argue that nested components, ordered by importance, can be extracted from pretrained models and selectively activated within the available computational budget. To this end, our proposed FlexRank method leverages low-rank weight decomposition with nested, importance-based consolidation to extract submodels of increasing capabilities. Our approach enables a "train-once, deploy-everywhere" paradigm offering a graceful trade-off between cost and performance without training from scratch for each budget - advancing practical deployment of large models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。