paper-with-me

Papers

WSM: Decay-Free Learning Rate Schedule via Checkpoint Merging for LLM Pre-training

2025-07-23 · Changxin Tian, Jiapeng Wang, Qian Zhao, Kunlong Chen, Jia Liu, Ziqi Liu, Jiaxin Mao, Wayne Xin Zhao, Zhiqiang Zhang, Jun Zhou arxiv

Recent advances in learning rate (LR) scheduling have demonstrated the effectiveness of decay-free approaches that eliminate the traditional decay phase while maintaining competitive performance. Model merging techniques have emerged as particularly promising solutions in this domain. We present Warmup-Stable and Merge (WSM), a general framework that establishes a formal connection between learning rate decay and model merging. WSM provides a unified theoretical foundation for emulating various decay strategies-including cosine decay, linear decay and inverse square root decay-as principled model averaging schemes, while remaining fully compatible with diverse optimization methods. Through extensive experiments, we identify merge duration-the training window for checkpoint aggregation-as the most critical factor influencing model performance, surpassing the importance of both checkpoint interval and merge quantity. Our framework consistently outperforms the widely-adopted Warmup-Stable-Decay (WSD) approach across multiple benchmarks, achieving significant improvements of +3.5% on MATH, +2.9% on HumanEval, and +5.5% on MMLU-Pro. The performance advantages extend to supervised fine-tuning scenarios, highlighting WSM's potential for long-term model refinement.

📄 PDF Abstract BibTeX arXiv:2507.17634

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

ScheduleFree+: Scaling Learning-Rate-Free & Schedule-Free Learning to Large Language Models

2026-05-18 · Aaron Defazio arxiv

Schedule-Free Learning has shown promise as a practical anytime training method for machine learning, showing success across dozens of standard benchmark problems. However, strong performance for LLM training has only be…

Anytime Training with Schedule-Free Spectral Optimization

2026-05-21 · Anuj Apte, Pranav Deshpande, Niraj Kumar, Shouvanik Chakrabarti 외 arxiv

Standard neural network training relies on learning-rate schedules tied to a fixed horizon, leading to strong path dependence and costly re-tuning as data availability changes. Schedule-Free (SF) methods address this by …

Continual Learning

Anytime Pretraining: Horizon-Free Learning-Rate Schedules with Weight Averaging

2026-02-03 · Alexandru Meterez, Pranav Ajit Nair, Depen Morwani, Cengiz Pehlevan 외 arxiv

Large language models are increasingly trained in continual or open-ended settings, where the total training horizon is not known in advance. Despite this, most existing pretraining recipes are not anytime: they rely on …

Understanding Warmup-Stable-Decay Learning Rates: A River Valley Loss Landscape Perspective

2024-10-07 · Kaiyue Wen, Zhiyuan Li, Jason Wang, David Hall 외

Training language models currently requires pre-determining a fixed compute budget because the typical cosine learning rate schedule depends on the total number of steps. In contrast, the Warmup-Stable-Decay (WSD) schedu…

When Flat Minima Fail: Characterizing INT4 Quantization Collapse After FP32 Convergence

2026-04-16 · Marcus Armstrong arxiv

Post-training quantization (PTQ) assumes that a well-converged model is a quantization-ready model. We show this assumption fails in a structured, measurable, and previously uncharacterized way. Using a calibration-free …