AI & ML
OpenGNS: An Open Dataset of the Gradient Noise Scale across Vision, Language, and Diffusion Models
arXiv cs.DC0 views
arXiv:2611.24331v1 Announce Type: cross Abstract: The Gradient Noise Scale (GNS) metric measures the signal-to-noise ratio of gradients in stochastic gradient descent and, as shown by McCandlish et al. (2018), can be used to approximate the Critical Batch Size (CBS) during neural network training. Despite its appearance in technical reports of frontier models, no open dataset of GNS measurements across modern architectures exists, limiting systematic study of the metric. We close this gap by releasing OpenGNS, a dataset of more than 400 GNS measurement trajectories spanning three domains: image classification with ResNets, autoregressive language modeling with Generative Pre-trained Transformers (GPTs), and image synthesis with Diffusion Transformers (DiTs), covering models from 11M to 1B parameters trained under Maximal Update Parameterization (muP) with dense batch-size and learning-rate sweeps. Building on this dataset, we study how the GNS depends on hyperparameters, scale, and training progress, and how it relates to the CBS. We find that the GNS behaves as a consistent "temperature of training", remaining stable across model sizes under muP at equivalent loss, and our results suggest task-dependent biases in its relation to the CBS: a tendency to overestimate the CBS for classification and to underestimate it for generative tasks, consistent in direction across common CBS definitions. The dataset is available at https://huggingface.co/datasets/olympique-marcel/OpenGNS
Source:https://arxiv.org/abs/2611.24331PreviousTraining-Free Domain Generation for One-Shot Federated Learning with Server-Side Pre-trained Diffusion ModelsNextDissecting an Unexplored SASS Control-Code Field: Barrier Exemption and Programmable Warp Priority
Related posts
90% of employers reject resumes that look like AI – yet half use AI to interview
0ZDNet
Old infrastructure, AI-driven fixes behind wave of US outages
0Phys.org Tech
Who gets Europe's scrap? Industry fights over future of waste exports
0Phys.org Tech

'The risks of AI spread faster': Report finds workers are handing over logins and even passwords to AI to make their jobs easier
0TechRadar

What employee adoption of Shadow AI is telling the enterprise
0TechRadar

Insta360's latest gadget is a pocket-sized color e-paper photo frame
0TechRadar