FlexMoE: Scaling Large-scale Sparse Pre-trained Model Training via Dynamic Device Placement

• URL: http://arxiv.org/abs/2304.03946v1
• Date: Sat, 8 Apr 2023 07:34:26 GMT
• Title: FlexMoE: Scaling Large-scale Sparse Pre-trained Model Training via Dynamic Device Placement
• Authors: Xiaonan Nie, Xupeng Miao, Zilong Wang, Zichao Yang, Jilong Xue, Lingxiao Ma, Gang Cao, Bin Cui
• Abstract summary: Mixture-of-Experts (MoEs) are becoming more popular and have demonstrated impressive pretraining scalability in various downstream tasks. MoEs are becoming a new data analytics paradigm in the data life cycle and suffering from unique challenges at scales, complexities, and granularities never before possible. In this paper, we propose a novel DNN training framework, FlexMoE, which systematically and transparently address the inefficiency caused by dynamic dataflow.
• Score: 19.639936387834677
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: With the increasing data volume, there is a trend of using large-scale pre-trained models to store the knowledge into an enormous number of model parameters. The training of these models is composed of lots of dense algebras, requiring a huge amount of hardware resources. Recently, sparsely-gated Mixture-of-Experts (MoEs) are becoming more popular and have demonstrated impressive pretraining scalability in various downstream tasks. However, such a sparse conditional computation may not be effective as expected in practical systems due to the routing imbalance and fluctuation problems. Generally, MoEs are becoming a new data analytics paradigm in the data life cycle and suffering from unique challenges at scales, complexities, and granularities never before possible. In this paper, we propose a novel DNN training framework, FlexMoE, which systematically and transparently address the inefficiency caused by dynamic dataflow. We first present an empirical analysis on the problems and opportunities of training MoE models, which motivates us to overcome the routing imbalance and fluctuation problems by a dynamic expert management and device placement mechanism. Then we introduce a novel scheduling module over the existing DNN runtime to monitor the data flow, make the scheduling plans, and dynamically adjust the model-to-hardware mapping guided by the real-time data traffic. A simple but efficient heuristic algorithm is exploited to dynamically optimize the device placement during training. We have conducted experiments on both NLP models (e.g., BERT and GPT) and vision models (e.g., Swin). And results show FlexMoE can achieve superior performance compared with existing systems on real-world workloads -- FlexMoE outperforms DeepSpeed by 1.70x on average and up to 2.10x, and outperforms FasterMoE by 1.30x on average and up to 1.45x.

Related papers

• Accelerating Large Language Model Pretraining via LFR Pedagogy: Learn, Focus, and Review [50.78587571704713]
  Large Language Model (LLM) pretraining traditionally relies on autoregressive language modeling on randomly sampled data blocks from web-scale datasets. We take inspiration from human learning techniques like spaced repetition to hypothesize that random data sampling for LLMs leads to high training cost and low quality models which tend to forget data. In order to effectively commit web-scale information to long-term memory, we propose the LFR (Learn, Focus, and Review) pedagogy.
  arXiv  Detail & Related papers  (2024-09-10T00:59:18Z)
• Dynamic Data Mixing Maximizes Instruction Tuning for Mixture-of-Experts [20.202031878825153]
  We propose a novel dynamic data mixture for MoE instruction tuning. Inspired by MoE's token routing preference, we build dataset-level representations and then capture the subtle differences among datasets. Results on two MoE models demonstrate the effectiveness of our approach on both downstream knowledge & reasoning tasks and open-ended queries.
  arXiv  Detail & Related papers  (2024-06-17T06:47:03Z)
• DynaMMo: Dynamic Model Merging for Efficient Class Incremental Learning for Medical Images [0.8213829427624407]
  Continual learning, the ability to acquire knowledge from new data while retaining previously learned information, is a fundamental challenge in machine learning. We propose Dynamic Model Merging, DynaMMo, a method that merges multiple networks at different stages of model training to achieve better computational efficiency. We evaluate DynaMMo on three publicly available datasets, demonstrating its effectiveness compared to existing approaches.
  arXiv  Detail & Related papers  (2024-04-22T11:37:35Z)
• DPOT: Auto-Regressive Denoising Operator Transformer for Large-Scale PDE Pre-Training [87.90342423839876]
  We present a new auto-regressive denoising pre-training strategy, which allows for more stable and efficient pre-training on PDE data. We train our PDE foundation model with up to 0.5B parameters on 10+ PDE datasets with more than 100k trajectories.
  arXiv  Detail & Related papers  (2024-03-06T08:38:34Z)
• Diffusion-Based Neural Network Weights Generation [80.89706112736353]
  D2NWG is a diffusion-based neural network weights generation technique that efficiently produces high-performing weights for transfer learning. Our method extends generative hyper-representation learning to recast the latent diffusion paradigm for neural network weights generation. Our approach is scalable to large architectures such as large language models (LLMs), overcoming the limitations of current parameter generation techniques.
  arXiv  Detail & Related papers  (2024-02-28T08:34:23Z)
• CoDBench: A Critical Evaluation of Data-driven Models for Continuous Dynamical Systems [8.410938527671341]
  We introduce CodBench, an exhaustive benchmarking suite comprising 11 state-of-the-art data-driven models for solving differential equations. Specifically, we evaluate 4 distinct categories of models, viz., feed forward neural networks, deep operator regression models, frequency-based neural operators, and transformer architectures. We conduct extensive experiments, assessing the operators' capabilities in learning, zero-shot super-resolution, data efficiency, robustness to noise, and computational efficiency.
  arXiv  Detail & Related papers  (2023-10-02T21:27:54Z)
• Asynchronous Multi-Model Dynamic Federated Learning over Wireless Networks: Theory, Modeling, and Optimization [20.741776617129208]
  Federated learning (FL) has emerged as a key technique for distributed machine learning (ML) We first formulate rectangular scheduling steps and functions to capture the impact of system parameters on learning performance. Our analysis sheds light on the joint impact of device training variables and asynchronous scheduling decisions.
  arXiv  Detail & Related papers  (2023-05-22T21:39:38Z)
• Online Evolutionary Neural Architecture Search for Multivariate Non-Stationary Time Series Forecasting [72.89994745876086]
  This work presents the Online Neuro-Evolution-based Neural Architecture Search (ONE-NAS) algorithm. ONE-NAS is a novel neural architecture search method capable of automatically designing and dynamically training recurrent neural networks (RNNs) for online forecasting tasks. Results demonstrate that ONE-NAS outperforms traditional statistical time series forecasting methods.
  arXiv  Detail & Related papers  (2023-02-20T22:25:47Z)
• Tutel: Adaptive Mixture-of-Experts at Scale [20.036168971435306]
  Sparsely-gated mixture-of-experts (MoE) has been widely adopted to scale deep learning models to trillion-plus parameters with fixed computational cost. We present Flex, a highly scalable stack design and implementation for MoE with dynamically adaptive parallelism and pipelining. Our evaluation shows that Flex efficiently and effectively runs a real-world MoE-based model named SwinV2-MoE, built upon Swin Transformer V2, a state-of-the-art computer vision architecture.
  arXiv  Detail & Related papers  (2022-06-07T15:20:20Z)
• Real-time Neural-MPC: Deep Learning Model Predictive Control for Quadrotors and Agile Robotic Platforms [59.03426963238452]
  We present Real-time Neural MPC, a framework to efficiently integrate large, complex neural network architectures as dynamics models within a model-predictive control pipeline. We show the feasibility of our framework on real-world problems by reducing the positional tracking error by up to 82% when compared to state-of-the-art MPC approaches without neural network dynamics.
  arXiv  Detail & Related papers  (2022-03-15T09:38:15Z)
• MoEfication: Conditional Computation of Transformer Models for Efficient Inference [66.56994436947441]
  Transformer-based pre-trained language models can achieve superior performance on most NLP tasks due to large parameter capacity, but also lead to huge computation cost. We explore to accelerate large-model inference by conditional computation based on the sparse activation phenomenon. We propose to transform a large model into its mixture-of-experts (MoE) version with equal model size, namely MoEfication.
  arXiv  Detail & Related papers  (2021-10-05T02:14:38Z)

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.