Power Scheduler: A Batch Size and Token Number Agnostic Learning Rate Scheduler

• URL: http://arxiv.org/abs/2408.13359v2
• Date: Wed, 11 Sep 2024 20:48:05 GMT
• Title: Power Scheduler: A Batch Size and Token Number Agnostic Learning Rate Scheduler
• Authors: Yikang Shen, Matthew Stallone, Mayank Mishra, Gaoyuan Zhang, Shawn Tan, Aditya Prasad, Adriana Meza Soria, David D. Cox, Rameswar Panda,
• Abstract summary: We study the correlation between optimal learning rate, batch size, and number of training tokens for the recently proposed WSD scheduler. We propose a new learning rate scheduler, Power scheduler, that is agnostic about the number of training tokens and batch size. Our 3B dense and MoE models trained with the Power scheduler achieve comparable performance as state-of-the-art small language models.
• Score: 34.416299887009195
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Finding the optimal learning rate for language model pretraining is a challenging task. This is not only because there is a complicated correlation between learning rate, batch size, number of training tokens, model size, and other hyperparameters but also because it is prohibitively expensive to perform a hyperparameter search for large language models with Billions or Trillions of parameters. Recent studies propose using small proxy models and small corpus to perform hyperparameter searches and transposing the optimal parameters to large models and large corpus. While the zero-shot transferability is theoretically and empirically proven for model size related hyperparameters, like depth and width, the zero-shot transfer from small corpus to large corpus is underexplored. In this paper, we study the correlation between optimal learning rate, batch size, and number of training tokens for the recently proposed WSD scheduler. After thousands of small experiments, we found a power-law relationship between variables and demonstrated its transferability across model sizes. Based on the observation, we propose a new learning rate scheduler, Power scheduler, that is agnostic about the number of training tokens and batch size. The experiment shows that combining the Power scheduler with Maximum Update Parameterization (muP) can consistently achieve impressive performance with one set of hyperparameters regardless of the number of training tokens, batch size, model size, and even model architecture. Our 3B dense and MoE models trained with the Power scheduler achieve comparable performance as state-of-the-art small language models. We open-source these pretrained models at https://ibm.biz/BdKhLa.

Related papers

• Scaling Smart: Accelerating Large Language Model Pre-training with Small Model Initialization [22.90653167145603]
  We introduce HyperCloning, a method that can expand the parameters of a pre-trained language model to those of a larger model with increased hidden dimensions. As a result, the larger model already inherits the predictive power and accuracy of the smaller model before the training starts.
  arXiv  Detail & Related papers  (2024-09-19T16:50:26Z)
• RepCNN: Micro-sized, Mighty Models for Wakeword Detection [3.4888176891918654]
  Always-on machine learning models require a very low memory and compute footprint. We show that a small convolutional model can be better trained by first its computation into a larger multi-branched architecture. We show that our always-on wake-word detector model, RepCNN, provides a good trade-off between latency and accuracy during inference.
  arXiv  Detail & Related papers  (2024-06-04T16:14:19Z)
• The Languini Kitchen: Enabling Language Modelling Research at Different Scales of Compute [66.84421705029624]
  We introduce an experimental protocol that enables model comparisons based on equivalent compute, measured in accelerator hours. We pre-process an existing large, diverse, and high-quality dataset of books that surpasses existing academic benchmarks in quality, diversity, and document length. This work also provides two baseline models: a feed-forward model derived from the GPT-2 architecture and a recurrent model in the form of a novel LSTM with ten-fold throughput.
  arXiv  Detail & Related papers  (2023-09-20T10:31:17Z)
• Pruning Large Language Models via Accuracy Predictor [0.0]
  Large language models (LLMs) containing tens of billions of parameters (or even more) have demonstrated impressive capabilities in various NLP tasks. We propose a novel pruning approach: firstly, a training set of a certain number of architecture-accuracy pairs is established, and then a non-neural model is trained as an accuracy predictor.
  arXiv  Detail & Related papers  (2023-09-18T06:38:24Z)
• Contrastive Alignment of Vision to Language Through Parameter-Efficient Transfer Learning [60.26952378997713]
  Contrastive vision-language models (e.g. CLIP) are created by updating all the parameters of a vision model and language model through contrastive training. We show that a minimal set of parameter updates ($$7%) can achieve the same performance as full-model training. We describe a series of experiments: we show that existing knowledge is conserved more strongly in parameter-efficient training.
  arXiv  Detail & Related papers  (2023-03-21T14:12:08Z)
• Zero-Shot Learners for Natural Language Understanding via a Unified Multiple Choice Perspective [26.41585967095811]
  Zero-shot learning aims to train a model on a given task such that it can address new learning tasks without any additional training. Our approach converts zero-shot learning into multiple-choice tasks, avoiding problems in commonly used large-scale generative models such as FLAN. Our approach shows state-of-the-art performance on several benchmarks and produces satisfactory results on tasks such as natural language inference and text classification.
  arXiv  Detail & Related papers  (2022-10-16T17:24:06Z)
• Zemi: Learning Zero-Shot Semi-Parametric Language Models from Multiple Tasks [77.90900650816046]
  We introduce $textZemi$, a zero-shot semi-parametric language model. We train $textZemi$ with a novel semi-parametric multitask prompted training paradigm. Specifically, we augment the multitask training and zero-shot evaluation with retrieval from a large-scale task-agnostic unlabeled corpus.
  arXiv  Detail & Related papers  (2022-10-01T04:08:50Z)
• Parameter-Efficient Sparsity for Large Language Models Fine-Tuning [63.321205487234074]
  We propose a. sparse-efficient Sparse Training (PST) method to reduce the number of trainable parameters during sparse-aware training. Experiments with diverse networks (i.e., BERT, RoBERTa and GPT-2) demonstrate PST performs on par or better than previous sparsity methods.
  arXiv  Detail & Related papers  (2022-05-23T02:43:45Z)
• Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning [81.3514358542452]
  Few-shot in-context learning (ICL) incurs substantial computational, memory, and storage costs because it involves processing all of the training examples every time a prediction is made. parameter-efficient fine-tuning offers an alternative paradigm where a small set of parameters are trained to enable a model to perform the new task. In this paper, we rigorously compare few-shot ICL and parameter-efficient fine-tuning and demonstrate that the latter offers better accuracy as well as dramatically lower computational costs.
  arXiv  Detail & Related papers  (2022-05-11T17:10:41Z)
• It's the Best Only When It Fits You Most: Finding Related Models for Serving Based on Dynamic Locality Sensitive Hashing [1.581913948762905]
  Preparation of training data is often a bottleneck in the lifecycle of deploying a deep learning model for production or research. This paper proposes an end-to-end process of searching related models for serving based on the similarity of the target dataset and the training datasets of the available models.
  arXiv  Detail & Related papers  (2020-10-13T22:52:13Z)
• The Right Tool for the Job: Matching Model and Instance Complexities [62.95183777679024]
  As NLP models become larger, executing a trained model requires significant computational resources incurring monetary and environmental costs. We propose a modification to contextual representation fine-tuning which, during inference, allows for an early (and fast) "exit" We test our proposed modification on five different datasets in two tasks: three text classification datasets and two natural language inference benchmarks.
  arXiv  Detail & Related papers  (2020-04-16T04:28:08Z)

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.