The Journey Matters: Average Parameter Count over Pre-training Unifies Sparse and Dense Scaling Laws

• URL: http://arxiv.org/abs/2501.12486v2
• Date: Sat, 15 Mar 2025 17:31:09 GMT
• Title: The Journey Matters: Average Parameter Count over Pre-training Unifies Sparse and Dense Scaling Laws
• Authors: Tian Jin, Ahmed Imtiaz Humayun, Utku Evci, Suvinay Subramanian, Amir Yazdanbakhsh, Dan Alistarh, Gintare Karolina Dziugaite,
• Abstract summary: We present the first systematic exploration of optimal sparse pre-training configurations for large language models.<n>We find that initiating pruning at 25% of total training compute and concluding at 75% achieves near-optimal final evaluation loss.<n>We propose a new scaling law that modifies the Chinchilla scaling law to use the average parameter count over pre-training.
• Score: 51.608402959163925
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Pruning eliminates unnecessary parameters in neural networks; it offers a promising solution to the growing computational demands of large language models (LLMs). While many focus on post-training pruning, sparse pre-training--which combines pruning and pre-training into a single phase--provides a simpler alternative. In this work, we present the first systematic exploration of optimal sparse pre-training configurations for LLMs through an examination of 80 unique pruning schedules across different sparsity levels and training durations. We find that initiating pruning at 25% of total training compute and concluding at 75% achieves near-optimal final evaluation loss. These findings provide valuable insights for efficient and effective sparse pre-training of LLMs. Furthermore, we propose a new scaling law that modifies the Chinchilla scaling law to use the average parameter count over pre-training. Through empirical and theoretical validation, we demonstrate that this modified scaling law accurately models evaluation loss for both sparsely and densely pre-trained LLMs, unifying scaling laws across pre-training paradigms. Our findings indicate that while sparse pre-training achieves the same final model quality as dense pre-training for equivalent compute budgets, it provides substantial benefits through reduced model size, enabling significant potential computational savings during inference.

Related papers

• Can Pre-training Indicators Reliably Predict Fine-tuning Outcomes of LLMs? [32.04523360747506]
  We construct a dataset using 50 1B parameter LLM variants with systematically varied pre-training configurations. We introduce novel unsupervised and supervised proxy metrics derived from pre-training that successfully reduce the relative performance prediction error rate by over 50%.
  arXiv  Detail & Related papers  (2025-04-16T21:19:09Z)
• The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models [69.798277882245]
  We introduce Unsupervised Prefix Fine-Tuning (UPFT) to enhance large language models' reasoning efficiency. UPFT removes the need for labeled data or exhaustive sampling. Experiments show that UPFT matches the performance of supervised methods.
  arXiv  Detail & Related papers  (2025-03-04T18:56:03Z)
• Unveiling Downstream Performance Scaling of LLMs: A Clustering-Based Perspective [5.09611816929943]
  Accurately predicting downstream task performance prior to model training is crucial for efficient resource allocation. Existing performance prediction methods suffer from limited accuracy and reliability. We propose a Clustering-On-Difficulty (COD) downstream performance prediction framework.
  arXiv  Detail & Related papers  (2025-02-24T15:44:57Z)
• P$^2$ Law: Scaling Law for Post-Training After Model Pruning [25.07013858614455]
  Pruning has become a widely adopted technique for reducing the hardware requirements of large language models (LLMs)<n>To recover model performance after pruning, post-training is commonly employed to mitigate the resulting performance degradation.<n>To balance post-training cost and model performance, it is necessary to explore the optimal amount of post-training data.
  arXiv  Detail & Related papers  (2024-11-15T15:28:42Z)
• Scaling Laws for Precision [73.24325358259753]
  We devise "precision-aware" scaling laws for both training and inference.<n>For inference, we find that the degradation introduced by post-training quantization increases as models are trained on more data.<n>For training, our scaling laws allow us to predict the loss of a model with different parts in different precisions.
  arXiv  Detail & Related papers  (2024-11-07T00:10:10Z)
• Scaling Laws for Predicting Downstream Performance in LLMs [75.28559015477137]
  This work focuses on the pre-training loss as a more-efficient metric for performance estimation. We extend the power law analytical function to predict domain-specific pre-training loss based on FLOPs across data sources. We employ a two-layer neural network to model the non-linear relationship between multiple domain-specific loss and downstream performance.
  arXiv  Detail & Related papers  (2024-10-11T04:57:48Z)
• Efficient Continual Pre-training by Mitigating the Stability Gap [68.49269649759005]
  We study the behavior of Large Language Models (LLMs) during continual pre-training. We propose three effective strategies to enhance LLM performance within a fixed compute budget. Our strategies improve the average medical task performance of the OpenLlama-3B model from 36.2% to 40.7% with only 40% of the original training budget.
  arXiv  Detail & Related papers  (2024-06-21T02:28:37Z)
• Temporal Scaling Law for Large Language Models [24.12384260752973]
  We propose the novel concept of Temporal Scaling Law, studying how the test loss of an LLM evolves as the training steps scale up. In contrast to modeling the test loss as a whole in a coarse-grained manner, we break it down and dive into the fine-grained test loss of each token position. We derive the much more precise temporal scaling law by studying the temporal patterns of the parameters in the dynamic hyperbolic-law.
  arXiv  Detail & Related papers  (2024-04-27T05:49:11Z)
• How Many Pretraining Tasks Are Needed for In-Context Learning of Linear Regression? [92.90857135952231]
  Transformers pretrained on diverse tasks exhibit remarkable in-context learning (ICL) capabilities. We study ICL in one of its simplest setups: pretraining a linearly parameterized single-layer linear attention model for linear regression.
  arXiv  Detail & Related papers  (2023-10-12T15:01:43Z)
• Prospect Pruning: Finding Trainable Weights at Initialization using Meta-Gradients [36.078414964088196]
  Pruning neural networks at initialization would enable us to find sparse models that retain the accuracy of the original network. Current methods are insufficient to enable this optimization and lead to a large degradation in model performance. We propose Prospect Pruning (ProsPr), which uses meta-gradients through the first few steps of optimization to determine which weights to prune. Our method achieves state-of-the-art pruning performance on a variety of vision classification tasks, with less data and in a single shot compared to existing pruning-at-initialization methods.
  arXiv  Detail & Related papers  (2022-02-16T15:18:55Z)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)

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.