Taming Transformer Without Using Learning Rate Warmup

• URL: http://arxiv.org/abs/2505.21910v1
• Date: Wed, 28 May 2025 02:55:28 GMT
• Title: Taming Transformer Without Using Learning Rate Warmup
• Authors: Xianbiao Qi, Yelin He, Jiaquan Ye, Chun-Guang Li, Bojia Zi, Xili Dai, Qin Zou, Rong Xiao,
• Abstract summary: Scaling Transformer to a large scale without using some technical tricks such as learning rate warump is an extremely challenging task.<n>We present a novel optimization strategy, ie, making the weight updating in successive steps smooth.<n>We conduct extensive experiments using ViT, Swin-Transformer and GPT, showing that our strategy can effectively and stably train these Transformers without using learning rate warmup.
• Score: 11.9495483265072
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Scaling Transformer to a large scale without using some technical tricks such as learning rate warump and using an obviously lower learning rate is an extremely challenging task, and is increasingly gaining more attention. In this paper, we provide a theoretical analysis for the process of training Transformer and reveal the rationale behind the model crash phenomenon in the training process, termed \textit{spectral energy concentration} of ${\bW_q}^{\top} \bW_k$, which is the reason for a malignant entropy collapse, where ${\bW_q}$ and $\bW_k$ are the projection matrices for the query and the key in Transformer, respectively. To remedy this problem, motivated by \textit{Weyl's Inequality}, we present a novel optimization strategy, \ie, making the weight updating in successive steps smooth -- if the ratio $\frac{\sigma_{1}(\nabla \bW_t)}{\sigma_{1}(\bW_{t-1})}$ is larger than a threshold, we will automatically bound the learning rate to a weighted multiple of $\frac{\sigma_{1}(\bW_{t-1})}{\sigma_{1}(\nabla \bW_t)}$, where $\nabla \bW_t$ is the updating quantity in step $t$. Such an optimization strategy can prevent spectral energy concentration to only a few directions, and thus can avoid malignant entropy collapse which will trigger the model crash. We conduct extensive experiments using ViT, Swin-Transformer and GPT, showing that our optimization strategy can effectively and stably train these Transformers without using learning rate warmup.

Related papers

• Provable Failure of Language Models in Learning Majority Boolean Logic via Gradient Descent [15.291830857281015]
  We investigate whether Transformers can truly learn simple majority functions when trained using gradient-based methods.<n>Our analysis demonstrates that even after $mathrmpoly(d)$ gradient queries, the generalization error of the Transformer model still remains substantially large.
  arXiv  Detail & Related papers  (2025-04-07T03:08:12Z)
• Benefits of Learning Rate Annealing for Tuning-Robustness in Stochastic Optimization [29.174036532175855]
  Learning rate in gradient methods is a critical hyperspecification that is notoriously costly to tune via standard grid search.<n>We identify a theoretical advantage of learning rate annealing schemes that decay the learning rate to zero at a rate, such as the widely-used cosine schedule.
  arXiv  Detail & Related papers  (2025-03-12T14:06:34Z)
• SAPPHIRE: Preconditioned Stochastic Variance Reduction for Faster Large-Scale Statistical Learning [18.055120576191204]
  Ill-conditioned objectives and nonsmooth regularizers undermine the performance of traditional convex methods.<n>We propose a variance-free solution for ill-conditioned composite large-scale machine learning problems.
  arXiv  Detail & Related papers  (2025-01-27T10:36:45Z)
• On the Role of Depth and Looping for In-Context Learning with Task Diversity [69.4145579827826]
  We study in-context learning for linear regression with diverse tasks. We show that multilayer Transformers are not robust to even distributional shifts as small as $O(e-L)$ in Wasserstein distance.
  arXiv  Detail & Related papers  (2024-10-29T03:27:56Z)
• In-context Learning for Mixture of Linear Regressions: Existence, Generalization and Training Dynamics [34.458004744956334]
  We prove that there exists a transformer capable of achieving a prediction error of order $mathcalO(sqrtd/n)$ with high probability.<n>We also analyze the training dynamics of transformers with single linear self-attention layers, demonstrating that, with appropriately parameters, gradient flow optimization over the population mean square loss converges to a global optimum.
  arXiv  Detail & Related papers  (2024-10-18T05:28:47Z)
• Can Looped Transformers Learn to Implement Multi-step Gradient Descent for In-context Learning? [69.4145579827826]
  We show a fast flow on the regression loss despite the gradient non-ity algorithms for our convergence landscape. This is the first theoretical analysis for multi-layer Transformer in this setting.
  arXiv  Detail & Related papers  (2024-10-10T18:29:05Z)
• On Mesa-Optimization in Autoregressively Trained Transformers: Emergence and Capability [34.43255978863601]
  Several suggest that transformers learn a mesa-optimizer during autorere training. We show that a stronger assumption related to the moments of data is the sufficient necessary condition that the learned mesa-optimizer can perform.
  arXiv  Detail & Related papers  (2024-05-27T05:41:06Z)
• Towards a Theoretical Understanding of the 'Reversal Curse' via Training Dynamics [45.69328374321502]
  Auto-regressive large language models (LLMs) show impressive capacities to solve many complex reasoning tasks. LLMs fail to conclude '$B gets A$' during inference even if the two sentences are semantically identical. We theoretically analyze the reversal curse via the training dynamics of gradient descent for two auto-regressive models.
  arXiv  Detail & Related papers  (2024-05-07T21:03:51Z)
• How do Transformers perform In-Context Autoregressive Learning? [76.18489638049545]
  We train a Transformer model on a simple next token prediction task. We show how a trained Transformer predicts the next token by first learning $W$ in-context, then applying a prediction mapping.
  arXiv  Detail & Related papers  (2024-02-08T16:24:44Z)
• p-Laplacian Transformer [7.2541371193810384]
  $p$-Laplacian regularization, rooted in graph and image signal processing, introduces a parameter $p$ to control the regularization effect on these data. We first show that the self-attention mechanism obtains the minimal Laplacian regularization. We then propose a novel class of transformers, namely the $p$-Laplacian Transformer (p-LaT)
  arXiv  Detail & Related papers  (2023-11-06T16:25:56Z)
• Modified Step Size for Enhanced Stochastic Gradient Descent: Convergence and Experiments [0.0]
  This paper introduces a novel approach to the performance of the gradient descent (SGD) algorithm by incorporating a modified decay step size based on $frac1sqrttt. The proposed step size integrates a logarithmic step term, leading to the selection of smaller values in the final iteration. To the effectiveness of our approach, we conducted numerical experiments on image classification tasks using the FashionMNIST, andARAR datasets.
  arXiv  Detail & Related papers  (2023-09-03T19:21:59Z)
• Transformers as Support Vector Machines [54.642793677472724]
  We establish a formal equivalence between the optimization geometry of self-attention and a hard-margin SVM problem. We characterize the implicit bias of 1-layer transformers optimized with gradient descent. We believe these findings inspire the interpretation of transformers as a hierarchy of SVMs that separates and selects optimal tokens.
  arXiv  Detail & Related papers  (2023-08-31T17:57:50Z)
• Stabilizing Transformer Training by Preventing Attention Entropy Collapse [56.45313891694746]
  We investigate the training dynamics of Transformers by examining the evolution of the attention layers. We show that $sigma$Reparam successfully prevents entropy collapse in the attention layers, promoting more stable training. We conduct experiments with $sigma$Reparam on image classification, image self-supervised learning, machine translation, speech recognition, and language modeling tasks.
  arXiv  Detail & Related papers  (2023-03-11T03:30:47Z)
• Revisiting Weighted Strategy for Non-stationary Parametric Bandits [82.1942459195896]
  This paper revisits the weighted strategy for non-stationary parametric bandits. We propose a refined analysis framework, which produces a simpler weight-based algorithm. Our new framework can be used to improve regret bounds of other parametric bandits.
  arXiv  Detail & Related papers  (2023-03-05T15:11:14Z)
• Understanding the Difficulty of Training Transformers [120.99980924577787]
  We show that unbalanced gradients are not the root cause of the instability of training. We propose Admin to stabilize the early stage's training and unleash its full potential in the late stage.
  arXiv  Detail & Related papers  (2020-04-17T13:59:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.