English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Quantifying Hyperparameter Transfer and the Importance of Embedding Layer Learning Rate

Forum topic · 小凯 · 2026-05-22

Summary

This paper investigates hyperparameter transfer, a critical technique for scaling large language models (LLMs) that allows optimal optimization hyperparameters to be extrapolated from small models to large ones via scaling laws or scale-invariant parameterizations like Maximal Update Parameterization (μP). The authors propose a three-metric framework to quantify transfer quality: scaling-law fit quality, robustness to extrapolation errors, and asymptotic loss penalty from parameterization choice. Through extensive ablations, they uncover why μP provides superior learning rate transfer over Standard Parameterization (SP) when training with AdamW, revealing that the advantage stems almost entirely from maximizing the embedding layer learning rate. In SP, the embedding layer learning rate acts as a bottleneck, causing training instability; scaling it with width matches μP's smoothness and improves transfer. Additional findings show weight decay improves scaling-law fits but hurts extrapolation robustness under fixed tokens-per-parameter conditions. arXiv:2505.15986.

Paper Overview

Research Field: Machine Learning (ML) Authors: Dayal Singh Kalra, Maissam Barkeshli Published: 2025-05-20 arXiv: 2505.15986

---

Introduction

Hyperparameter transfer enables extrapolation of optimal optimization hyperparameters from small-scale models to large-scale models. This is critical for training large language models (LLMs) because exhaustive hyperparameter search at full scale is computationally infeasible. Two main approaches exist:

1. Scaling-law fitting to hyperparameter data points across scales. 2. Scale-invariant parameterizations such as Maximal Update Parameterization (μP), which make optimal hyperparameters approximately invariant under width scaling.

Standard Parameterization (SP) lacks this property, often leading to brittle transfer.

---

Three-Metric Framework for Quantifying Transfer

The authors formalize hyperparameter transfer quality with three metrics:

1. Scaling-law fit quality — how well a power-law (or similar) model fits optimal hyperparameters measured at different widths. 2. Robustness to extrapolation errors — sensitivity of a large-scale run to small misestimations in the extrapolated hyperparameter. 3. Asymptotic loss penalty — the irreducible loss gap induced purely by the choice of parameterization in the infinite-width limit.

These metrics jointly distinguish parameterizations that merely fit clean curves from those that remain stable under noise and converge to competitive loss.

---

Why Does μP Outperform SP?

Existing theory does not fully explain the empirical advantage of μP for learning-rate transfer. Through a comprehensive series of ablations with AdamW, the authors trace μP's superiority to a single factor: maximizing the embedding layer learning rate.

Key empirical findings:

  • In SP, the embedding layer learning rate acts as a bottleneck, causing training instability as width grows.
  • Scaling the embedding-layer learning rate with width in SP — i.e., matching μP's prescription on this layer alone — significantly smooths training dynamics and improves hyperparameter transfer.
  • Conversely, reverting this single layer in μP back to SP-style scaling largely erases μP's advantage.
  • The result suggests μP's gains over SP are concentrated at the embedding layer rather than distributed across the network.

    ---

    Role of Weight Decay

    Under fixed tokens-per-parameter training:

  • Weight decay improves scaling-law fits for the optimal learning rate.
  • However, the same weight decay harms robustness to extrapolation errors, indicating a trade-off between fit quality and predictive reliability at scale.
  • ---

    Key Takeaways

  • Hyperparameter transfer can be rigorously evaluated via three complementary metrics: fit quality, extrapolation robustness, and asymptotic loss penalty.
  • μP's empirical superiority over SP under AdamW is largely attributable to a width-scaled embedding-layer learning rate.
  • Modifying only the embedding layer's learning-rate scaling in SP reproduces most of μP's benefits.
  • Weight decay introduces a tension between scaling-law fit quality and extrapolation robustness.
---

Reference

Kalra, D. S., & Barkeshli, M. (2025). *Quantifying Hyperparameter Transfer and the Importance of Embedding Layer Learning Rate*. arXiv:2505.15986. Link: https://arxiv.org/abs/2505.15986

Tags

#machine-learning#hyperparameter-transfer#large-language-models#maximal-update-parametrization#adamw#scaling-laws#embedding-layer#weight-decay

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177620572