Stochastic Rounding for LLM Training: Theory and Practice

• URL: http://arxiv.org/abs/2502.20566v1
• Date: Thu, 27 Feb 2025 22:08:08 GMT
• Title: Stochastic Rounding for LLM Training: Theory and Practice
• Authors: Kaan Ozkara, Tao Yu, Youngsuk Park,
• Abstract summary: We leverage rounding (SR) to address numerical errors with low-precision representation.<n> Empirical results from pre-training models with up to 6.7B parameters, for the first time, demonstrate that our BF16 with SR strategy outperforms (BF16, FP32) mixed precision strategies.
• Score: 15.071158535119539
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As the parameters of Large Language Models (LLMs) have scaled to hundreds of billions, the demand for efficient training methods -- balancing faster computation and reduced memory usage without sacrificing accuracy -- has become more critical than ever. In recent years, various mixed precision strategies, which involve different precision levels for optimization components, have been proposed to increase training speed with minimal accuracy degradation. However, these strategies often require manual adjustments and lack theoretical justification. In this work, we leverage stochastic rounding (SR) to address numerical errors of training with low-precision representation. We provide theoretical analyses of implicit regularization and convergence under the Adam optimizer when SR is utilized. With the insights from these analyses, we extend previous BF16 + SR strategy to be used in distributed settings, enhancing the stability and performance for large scale training. Empirical results from pre-training models with up to 6.7B parameters, for the first time, demonstrate that our BF16 with SR strategy outperforms (BF16, FP32) mixed precision strategies, achieving better validation perplexity, up to $1.54\times$ higher throughput, and $30\%$ less memory usage.

Related papers

• The Journey Matters: Average Parameter Count over Pre-training Unifies Sparse and Dense Scaling Laws [51.608402959163925]
  We present the first systematic exploration of optimal sparse pre-training configurations for large language models. We find that initiating pruning at 25% of total training compute and concluding at 75% achieves near-optimal final evaluation loss. We propose a new scaling law that modifies the Chinchilla scaling law to use the average parameter count over pre-training.
  arXiv  Detail & Related papers  (2025-01-21T20:23:22Z)
• HALO: Hadamard-Assisted Lower-Precision Optimization for LLMs [45.37278584462772]
  We present HALO, a novel quantization-aware training approach for Transformers.<n>Our approach ensures that all large matrix multiplications during the forward and backward passes are executed in lower precision.<n>Applying to LLAMA-family models, HALO achieves near-full-precision-equivalent results during fine-tuning on various tasks.
  arXiv  Detail & Related papers  (2025-01-05T18:41:54Z)
• Optimization Hyper-parameter Laws for Large Language Models [52.49860340549727]
  We present Opt-Laws, a framework that captures the relationship between hyper- parameters and training outcomes.<n>Our validation across diverse model sizes and data scales demonstrates Opt-Laws' ability to accurately predict training loss.<n>This approach significantly reduces computational costs while enhancing overall model performance.
  arXiv  Detail & Related papers  (2024-09-07T09:37:19Z)
• Enhancing One-shot Pruned Pre-trained Language Models through Sparse-Dense-Sparse Mechanism [25.36736897890854]
  Pre-trained language models (PLMs) are engineered to be robust in contextual understanding and exhibit outstanding performance in various natural language processing tasks. Modern pruning strategies employ one-shot techniques to compress PLMs without the need for retraining on task-specific or otherwise general data. We propose SDS, a Sparse-Dense-Sparse pruning framework to enhance the performance of the pruned PLMs from a weight distribution optimization perspective.
  arXiv  Detail & Related papers  (2024-08-20T01:05:45Z)
• Pruning Large Language Models with Semi-Structural Adaptive Sparse Training [17.381160429641316]
  Adaptive Sparse Trainer (AST) is a novel and efficient retraining framework tailored for semi-structured sparse models.<n>AST reduces the perplexity and zero-shot accuracy gap between dense and 2:4 semi-structured sparse models to 0.6 and 1.16%, respectively.
  arXiv  Detail & Related papers  (2024-07-30T06:33:44Z)
• To FP8 and Back Again: Quantifying Reduced Precision Effects on LLM Training Stability [7.115739465137031]
  BrainFloat16 (BF16) precision has become the de facto standard for large language model pretraining. However, prior experience with FP16, which was found to be less stable than BF16, raises concerns as to whether FP8 can be a cost-effective option for LLM training. We propose new evaluation techniques and a new metric for quantifying loss landscape sharpness in autoregressive language models.
  arXiv  Detail & Related papers  (2024-05-29T02:42:23Z)
• Zero-Shot Sharpness-Aware Quantization for Pre-trained Language Models [88.80146574509195]
  Quantization is a promising approach for reducing memory overhead and accelerating inference. We propose a novel-aware quantization (ZSAQ) framework for the zero-shot quantization of various PLMs.
  arXiv  Detail & Related papers  (2023-10-20T07:09:56Z)
• Surrogate Lagrangian Relaxation: A Path To Retrain-free Deep Neural Network Pruning [9.33753001494221]
  Network pruning is a widely used technique to reduce computation cost and model size for deep neural networks. In this paper, we develop a systematic weight-pruning optimization approach based on Surrogate Lagrangian relaxation.
  arXiv  Detail & Related papers  (2023-04-08T22:48:30Z)
• Dynamic Sparse Training via Balancing the Exploration-Exploitation Trade-off [19.230329532065635]
  Sparse training could significantly mitigate the training costs by reducing the model size. Existing sparse training methods mainly use either random-based or greedy-based drop-and-grow strategies. In this work, we consider the dynamic sparse training as a sparse connectivity search problem. Experimental results show that sparse models (up to 98% sparsity) obtained by our proposed method outperform the SOTA sparse training methods.
  arXiv  Detail & Related papers  (2022-11-30T01:22:25Z)
• Distributionally Robust Models with Parametric Likelihood Ratios [123.05074253513935]
  Three simple ideas allow us to train models with DRO using a broader class of parametric likelihood ratios. We find that models trained with the resulting parametric adversaries are consistently more robust to subpopulation shifts when compared to other DRO approaches.
  arXiv  Detail & Related papers  (2022-04-13T12:43:12Z)
• MEST: Accurate and Fast Memory-Economic Sparse Training Framework on the Edge [72.16021611888165]
  This paper proposes a novel Memory-Economic Sparse Training (MEST) framework targeting for accurate and fast execution on edge devices. The proposed MEST framework consists of enhancements by Elastic Mutation (EM) and Soft Memory Bound (&S) Our results suggest that unforgettable examples can be identified in-situ even during the dynamic exploration of sparsity masks.
  arXiv  Detail & Related papers  (2021-10-26T21:15:17Z)
• 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.