paper-with-me

Papers

SAGE: Sign-Adaptive Gradient for Memory-Efficient LLM Optimization

2026-04-09 · Wooin Lee, Hyun-Tae Kim arxiv

The AdamW optimizer, while standard for LLM pretraining, is a critical memory bottleneck, consuming optimizer states equivalent to twice the model's size. Although light-state optimizers like SinkGD attempt to address this issue, we identify the embedding layer dilemma: these methods fail to handle the sparse, high-variance gradients inherent to embeddings, forcing a hybrid design that reverts to AdamW and partially negates the memory gains. We propose SAGE (Sign Adaptive GradiEnt), a novel optimizer that resolves this dilemma by replacing AdamW in this hybrid structure. SAGE combines a Lion-style update direction with a new, memory-efficient $O(d)$ adaptive scale. This scale acts as a "safe damper," provably bounded by 1.0, which tames high-variance dimensions more effectively than existing methods. This superior stability allows SAGE to achieve better convergence. On Llama models up to 1.3B parameters, our SAGE-based hybrid achieves new state-of-the-art perplexity, outperforming all baselines, including SinkGD hybrid, while significantly reducing optimizer state memory.

📄 PDF Abstract BibTeX arXiv:2604.07663

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

CAME: Confidence-guided Adaptive Memory Efficient Optimization

2023-07-05 · Yang Luo, Xiaozhe Ren, Zangwei Zheng, Zhuo Jiang 외

Adaptive gradient methods, such as Adam and LAMB, have demonstrated excellent performance in the training of large language models. Nevertheless, the need for adaptivity requires maintaining second-moment estimates of th…

Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients

2024-07-11 · Zhenyu Zhang, Ajay Jaiswal, Lu Yin, Shiwei Liu 외

Training Large Language Models (LLMs) is memory-intensive due to the large number of parameters and associated optimization states. GaLore, a recent method, reduces memory usage by projecting weight gradients into a low-…

Quantization

Tri-Accel: Curvature-Aware Precision-Adaptive and Memory-Elastic Optimization for Efficient GPU Usage

2025-08-23 · Mohsen Sheibanian, Pouya Shaeri, Alimohammad Beigi, Ryan T. Woo 외 arxiv

Deep neural networks are increasingly bottlenecked by the cost of optimization, both in terms of GPU memory and compute time. Existing acceleration techniques, such as mixed precision, second-order methods, and batch siz…

GRASS: Gradient-based Adaptive Layer-wise Importance Sampling for Memory-efficient Large Language Model Fine-tuning

2026-04-09 · Kaiyuan Tian, Yu Tang, Gongqingjian Jiang, Baihui Liu 외 arxiv

Full-parameter fine-tuning of large language models is constrained by substantial GPU memory requirements. Low-rank adaptation methods mitigate this challenge by updating only a subset of parameters. However, these appro…

Fine-tuning Quantized Neural Networks with Zeroth-order Optimization

2025-05-19 · Sifeng Shang, Jiayi Zhou, Chenyu Lin, Minxian Li 외

As the size of large language models grows exponentially, GPU memory has become a bottleneck for adapting these models to downstream tasks. In this paper, we aim to push the limits of memory-efficient training by minimiz…

GPUQuantization