paper-with-me

Papers

Training Dynamics of the Cooldown Stage in Warmup-Stable-Decay Learning Rate Scheduler

2025-08-02 · Aleksandr Dremov, Alexander Hägele, Atli Kosson, Martin Jaggi arxiv

Learning rate scheduling is essential in transformer training, where the final annealing plays a crucial role in getting the best performance. However, the mechanisms behind this cooldown phase, with its characteristic drop in loss, remain poorly understood. To address this, we provide a comprehensive analysis focusing solely on the cooldown phase in the Warmup-Stable-Decay (WSD) learning rate scheduler. Our analysis reveals that different cooldown shapes reveal a fundamental bias-variance trade-off in the resulting models, with shapes that balance exploration and exploitation consistently outperforming alternatives. Similarly, we find substantial performance variations $\unicode{x2013}$ comparable to those from cooldown shape selection $\unicode{x2013}$ when tuning AdamW hyperparameters. Notably, we observe consistent improvements with higher values of $β_2$ during cooldown. From a loss landscape perspective, we provide visualizations of the landscape during cooldown, supporting the river valley loss perspective empirically. These findings offer practical recommendations for configuring the WSD scheduler in transformer training, emphasizing the importance of optimizing the cooldown phase alongside traditional hyperparameter tuning.

📄 PDF Abstract BibTeX arXiv:2508.01483

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Same Loss, Same Noise, Opposite Schedules: Noise Structure and Optimizer Normalization Jointly Determine Whether Learning-Rate Cooldown Helps

2026-07-14 · Subham Singh, Ashutosh Mishra, Subha Raut arxiv

The cooldown phase of a warmup-stable-decay (WSD) learning-rate schedule, now a default in large-model pretraining, lowers the final training loss in some settings and does nothing in others. We give a provable account o…

When Will Gradient Regularization Be Harmful?

2024-06-14 · Yang Zhao, Hao Zhang, Xiuyuan Hu

Gradient regularization (GR), which aims to penalize the gradient norm atop the loss function, has shown promising results in training modern over-parameterized deep neural networks. However, can we trust this powerful t…

Universal Dynamics of Warmup Stable Decay: understanding WSD beyond Transformers

2026-01-13 · Annalisa Belloni, Lorenzo Noci, Antonio Orvieto arxiv

The Warmup Stable Decay (WSD) learning rate scheduler has recently become popular, largely due to its good performance and flexibility when training large language models. It remains an open question whether the remarkab…

Stable-DiffCoder: Pushing the Frontier of Code Diffusion Large Language Model

2026-01-22 · Chenghao Fan, Wen Heng, Bo Li, Sichen Liu 외 arxiv

Diffusion-based language models (DLLMs) offer non-sequential, block-wise generation and richer data reuse compared to autoregressive (AR) models, but existing code DLLMs still lag behind strong AR baselines under compara…

Continual PretrainingData Augmentation

Weight Decay may matter more than muP for Learning Rate Transfer in Practice

2025-10-21 · Atli Kosson, Jeremy Welborn, Yang Liu, Martin Jaggi 외 arxiv

Transferring the optimal learning rate from small to large neural networks can enable efficient training at scales where hyperparameter tuning is otherwise prohibitively expensive. To this end, the Maximal Update Paramet…