$μ$pscaling small models: Principled warm starts and hyperparameter transfer

• URL: http://arxiv.org/abs/2602.10545v1
• Date: Wed, 11 Feb 2026 05:37:22 GMT
• Title: $μ$pscaling small models: Principled warm starts and hyperparameter transfer
• Authors: Yuxin Ma, Nan Chen, Mateo Díaz, Soufiane Hayou, Dmitriy Kunisky, Soledad Villar,
• Abstract summary: We introduce a principled upscaling method applicable to a broad range of architectures and architectures.<n>We empirically demonstrate that this method is effective on realistic datasets and architectures.
• Score: 30.73409211009394
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern large-scale neural networks are often trained and released in multiple sizes to accommodate diverse inference budgets. To improve efficiency, recent work has explored model upscaling: initializing larger models from trained smaller ones in order to transfer knowledge and accelerate convergence. However, this method can be sensitive to hyperparameters that need to be tuned at the target upscaled model size, which is prohibitively costly to do directly. It remains unclear whether the most common workaround -- tuning on smaller models and extrapolating via hyperparameter scaling laws -- is still sound when using upscaling. We address this with principled approaches to upscaling with respect to model widths and efficiently tuning hyperparameters in this setting. First, motivated by $μ$P and any-dimensional architectures, we introduce a general upscaling method applicable to a broad range of architectures and optimizers, backed by theory guaranteeing that models are equivalent to their widened versions and allowing for rigorous analysis of infinite-width limits. Second, we extend the theory of $μ$Transfer to a hyperparameter transfer technique for models upscaled using our method and empirically demonstrate that this method is effective on realistic datasets and architectures.

Related papers

• Theoretical Foundations of Scaling Law in Familial Models [46.506708373314375]
  We introduce Granularity (G) as a fundamental scaling variable alongside model size (N) and training tokens (D)<n>Our results reveal that the granularity penalty follows a multiplicative power law with an extremely small exponent.<n> Practically, it validates the "train once, deploy many" paradigm, demonstrating that deployment flexibility is achievable.
  arXiv  Detail & Related papers  (2025-12-29T12:01:58Z)
• $μ$-Parametrization for Mixture of Experts [8.950722808523981]
  Mixture-of-Experts (MoE) are emerging as a leading architecture in extremely large models.<n>$mu$Transfer allows seamless transfer of optimal hyper parameters across model scales.<n>Experiments demonstrate that the optimal learning rate reliably transfers across model sizes.
  arXiv  Detail & Related papers  (2025-08-13T12:31:27Z)
• LESA: Learnable LLM Layer Scaling-Up [57.0510934286449]
  Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive.<n>Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger ones.<n>We propose textbfLESA, a novel learnable method for depth scaling-up.
  arXiv  Detail & Related papers  (2025-02-19T14:58:48Z)
• Merging Models on the Fly Without Retraining: A Sequential Approach to Scalable Continual Model Merging [75.93960998357812]
  Deep model merging represents an emerging research direction that combines multiple fine-tuned models to harness their capabilities across different tasks and domains.<n>Current model merging techniques focus on merging all available models simultaneously, with weight matrices-based methods being the predominant approaches.<n>We propose a training-free projection-based continual merging method that processes models sequentially.
  arXiv  Detail & Related papers  (2025-01-16T13:17:24Z)
• Warmstarting for Scaling Language Models [47.691182347349894]
  Scaling model sizes to scale performance has worked remarkably well for the current large language models paradigm. High training costs for contemporary scales of data and models result in a lack of thorough understanding of how to tune and arrive at such training setups. One direction to ameliorate the cost of pretraining large models is to warmstart the large-scale training from smaller models that are cheaper to tune.
  arXiv  Detail & Related papers  (2024-11-11T20:02:29Z)
• SaRA: High-Efficient Diffusion Model Fine-tuning with Progressive Sparse Low-Rank Adaptation [52.6922833948127]
  In this work, we investigate the importance of parameters in pre-trained diffusion models.<n>We propose a novel model fine-tuning method to make full use of these ineffective parameters.<n>Our method enhances the generative capabilities of pre-trained models in downstream applications.
  arXiv  Detail & Related papers  (2024-09-10T16:44:47Z)
• SMILE: Zero-Shot Sparse Mixture of Low-Rank Experts Construction From Pre-Trained Foundation Models [85.67096251281191]
  We present an innovative approach to model fusion called zero-shot Sparse MIxture of Low-rank Experts (SMILE) construction. SMILE allows for the upscaling of source models into an MoE model without extra data or further training. We conduct extensive experiments across diverse scenarios, such as image classification and text generation tasks, using full fine-tuning and LoRA fine-tuning.
  arXiv  Detail & Related papers  (2024-08-19T17:32:15Z)
• Majority Kernels: An Approach to Leverage Big Model Dynamics for Efficient Small Model Training [32.154166415680066]
  Methods like distillation, compression or quantization help leverage the highly performant large models to induce smaller performant ones. This paper explores the hypothesis that a single training run can simultaneously train a larger model for performance and derive a smaller model for deployment.
  arXiv  Detail & Related papers  (2024-02-07T17:07:41Z)
• Scaling Pre-trained Language Models to Deeper via Parameter-efficient Architecture [68.13678918660872]
  We design a more capable parameter-sharing architecture based on matrix product operator (MPO) MPO decomposition can reorganize and factorize the information of a parameter matrix into two parts. Our architecture shares the central tensor across all layers for reducing the model size.
  arXiv  Detail & Related papers  (2023-03-27T02:34:09Z)

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.