Enhancing Stability for Large Language Models Training in Constrained Bandwidth Networks

• URL: http://arxiv.org/abs/2407.01614v3
• Date: Sun, 06 Oct 2024 01:18:35 GMT
• Title: Enhancing Stability for Large Language Models Training in Constrained Bandwidth Networks
• Authors: Yun Dai, Tejas Dharamsi, Byron Hsu, Tao Song, Hamed Firooz,
• Abstract summary: We show how potential race conditions in the hierarchical partitioning (hpZ) scheme cause instability when training models with billions of parameters. We then propose a modification to the partitioning algorithm that addresses these convergence challenges while maintaining competitive training efficiency. The updated algorithm enables robust training of larger models with 98% throughput and model training speed improvement without sacrificing the quality of convergence.
• Score: 8.049237611207113
• License:
• Abstract: Training extremely large language models (LLMs) with billions of parameters is a computationally intensive task that pushes the limits of current data parallel training systems. While techniques like ZeRO++ have enabled efficient distributed training of such giant models on inexpensive low-bandwidth clusters, they can suffer from convergence issues due to potential race conditions in the hierarchical partitioning (hpZ) scheme employed to reduce cross-machine communication. In this work, we first show how these race conditions cause instability when training models with billions of parameters. We then propose a modification to the partitioning algorithm that addresses these convergence challenges while maintaining competitive training efficiency. Empirical evaluation on training the multi-billion parameters Falcon Models and Llama-2 models demonstrates the updated algorithm's ability to achieve reliable convergence on these massive models, where stock ZeRO++ hpZ fails to converge. The updated algorithm enables robust training of larger models with 98\% throughput and model training speed improvement without sacrificing the quality of convergence.

Related papers

• Transferable Post-training via Inverse Value Learning [83.75002867411263]
  We propose modeling changes at the logits level during post-training using a separate neural network (i.e., the value network) After training this network on a small base model using demonstrations, this network can be seamlessly integrated with other pre-trained models during inference. We demonstrate that the resulting value network has broad transferability across pre-trained models of different parameter sizes.
  arXiv  Detail & Related papers  (2024-10-28T13:48:43Z)
• A Multi-Level Framework for Accelerating Training Transformer Models [5.268960238774481]
  Training large-scale deep learning models poses an unprecedented demand for computing power. We propose a multi-level framework for training acceleration based on Coalescing, De-coalescing and Interpolation. We prove that the proposed framework reduces the computational cost by about 20% on training BERT/GPT-Base models and up to 51.6% on training the BERT-Large model.
  arXiv  Detail & Related papers  (2024-04-07T03:04:34Z)
• 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)
• Ravnest: Decentralized Asynchronous Training on Heterogeneous Devices [0.0]
  Ravnest facilitates decentralized training by efficiently organizing compute nodes into clusters. We have framed our asynchronous SGD loss function as a block structured optimization problem with delayed updates.
  arXiv  Detail & Related papers  (2024-01-03T13:07:07Z)
• Towards a Better Theoretical Understanding of Independent Subnetwork Training [56.24689348875711]
  We take a closer theoretical look at Independent Subnetwork Training (IST) IST is a recently proposed and highly effective technique for solving the aforementioned problems. We identify fundamental differences between IST and alternative approaches, such as distributed methods with compressed communication.
  arXiv  Detail & Related papers  (2023-06-28T18:14:22Z)
• One-stop Training of Multiple Capacity Models [74.87789190840527]
  We propose a novel one-stop training framework to jointly train high-capacity and low-capactiy models. Unlike knowledge distillation, where multiple capacity models are trained from scratch separately, our approach integrates supervisions from different capacity models simultaneously.
  arXiv  Detail & Related papers  (2023-05-23T13:44:09Z)
• SWARM Parallelism: Training Large Models Can Be Surprisingly Communication-Efficient [69.61083127540776]
  Deep learning applications benefit from using large models with billions of parameters. Training these models is notoriously expensive due to the need for specialized HPC clusters. We consider alternative setups for training large models: using cheap "preemptible" instances or pooling existing resources from multiple regions.
  arXiv  Detail & Related papers  (2023-01-27T18:55:19Z)
• Fast-Convergent Federated Learning [82.32029953209542]
  Federated learning is a promising solution for distributing machine learning tasks through modern networks of mobile devices. We propose a fast-convergent federated learning algorithm, called FOLB, which performs intelligent sampling of devices in each round of model training.
  arXiv  Detail & Related papers  (2020-07-26T14:37:51Z)
• Deep Generative Models that Solve PDEs: Distributed Computing for Training Large Data-Free Models [25.33147292369218]
  Recent progress in scientific machine learning (SciML) has opened up the possibility of training novel neural network architectures that solve complex partial differential equations (PDEs) Here we report on a software framework for data parallel distributed deep learning that resolves the twin challenges of training these large SciML models. Our framework provides several out of the box functionality including (a) loss integrity independent of number of processes, (b) synchronized batch normalization, and (c) distributed higher-order optimization methods.
  arXiv  Detail & Related papers  (2020-07-24T22:42:35Z)

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.