GEAR: A GPU-Centric Experience Replay System for Large Reinforcement Learning Models

• URL: http://arxiv.org/abs/2310.05205v1
• Date: Sun, 8 Oct 2023 15:39:43 GMT
• Title: GEAR: A GPU-Centric Experience Replay System for Large Reinforcement Learning Models
• Authors: Hanjing Wang, Man-Kit Sit, Congjie He, Ying Wen, Weinan Zhang, Jun Wang, Yaodong Yang, Luo Mai
• Abstract summary: GEAR is designed to perform scalable reinforcement learning (RL) with large sequence models (such as transformers) It is equipped with GPU kernels capable of collecting trajectories using zero-copy access to host memory, along with remote-directed-memory access over InfiniBand. Gear can achieve performance levels up to 6x greater than Reverb when training state-of-the-art large RL models.
• Score: 32.23853007467266
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: This paper introduces a distributed, GPU-centric experience replay system, GEAR, designed to perform scalable reinforcement learning (RL) with large sequence models (such as transformers). With such models, existing systems such as Reverb face considerable bottlenecks in memory, computation, and communication. GEAR, however, optimizes memory efficiency by enabling the memory resources on GPU servers (including host memory and device memory) to manage trajectory data. Furthermore, it facilitates decentralized GPU devices to expedite various trajectory selection strategies, circumventing computational bottlenecks. GEAR is equipped with GPU kernels capable of collecting trajectories using zero-copy access to host memory, along with remote-directed-memory access over InfiniBand, improving communication efficiency. Cluster experiments have shown that GEAR can achieve performance levels up to 6x greater than Reverb when training state-of-the-art large RL models. GEAR is open-sourced at https://github.com/bigrl-team/gear.

Related papers

• Mind the Memory Gap: Unveiling GPU Bottlenecks in Large-Batch LLM Inference [4.497936996651617]
  Large language models have been widely adopted across different tasks, but their auto-regressive generation nature often leads to inefficient resource utilization during inference. In this paper, through an in-depth GPU-level analysis, we reveal that large-batch inference remains memory-bound, with most GPU compute capabilities underutilized due to DRAM bandwidth saturation as the primary bottleneck.
  arXiv  Detail & Related papers  (2025-03-11T11:21:35Z)
• Scaling Large Language Model Training on Frontier with Low-Bandwidth Partitioning [2.685330831042324]
  We propose a collection of communication and optimization strategies for ZeRO++ to reduce communication costs and improve memory utilization. For a 20B GPT model, we observe a 1.71x increase in TFLOPS per GPU when compared with ZeRO++ up to 384 GCDs and a scaling efficiency of 0.94 for up to 384 GCDs.
  arXiv  Detail & Related papers  (2025-01-08T04:19:57Z)
• APOLLO: SGD-like Memory, AdamW-level Performance [61.53444035835778]
  Large language models (LLMs) are notoriously memory-intensive during training. Various memory-efficient Scals have been proposed to reduce memory usage. They face critical challenges: (i) costly SVD operations; (ii) significant performance trade-offs compared to AdamW; and (iii) still substantial memory overhead to maintain competitive performance.
  arXiv  Detail & Related papers  (2024-12-06T18:55:34Z)
• Memory-Efficient Training for Deep Speaker Embedding Learning in Speaker Verification [50.596077598766975]
  We explore a memory-efficient training strategy for deep speaker embedding learning in resource-constrained scenarios. For activations, we design two types of reversible neural networks which eliminate the need to store intermediate activations. For states, we introduce a dynamic quantization approach that replaces the original 32-bit floating-point values with a dynamic tree-based 8-bit data type.
  arXiv  Detail & Related papers  (2024-12-02T06:57:46Z)
• MoE-Lightning: High-Throughput MoE Inference on Memory-constrained GPUs [55.95879347182669]
  MoE architecture is renowned for its ability to increase model capacity without a proportional increase in inference cost. MoE-Lightning introduces a novel CPU-GPU-I/O pipelining schedule, CGOPipe, with paged weights to achieve high resource utilization. MoE-Lightning can achieve up to 10.3x higher throughput than state-of-the-art offloading-enabled LLM inference systems for Mixtral 8x7B on a single T4 GPU (16GB)
  arXiv  Detail & Related papers  (2024-11-18T01:06:12Z)
• FastGL: A GPU-Efficient Framework for Accelerating Sampling-Based GNN Training at Large Scale [29.272368697268433]
  Graph Neural Networks (GNNs) have shown great superiority on non-Euclidean graph data. We propose FastGL, a GPU-efficient Framework for accelerating sampling-based training of GNN at Large scale. FastGL can achieve an average speedup of 11.8x, 2.2x and 1.5x over the state-of-the-art frameworks PyG, DGL, and GNNLab, respectively.
  arXiv  Detail & Related papers  (2024-09-23T11:45:47Z)
• AI and Memory Wall [81.06494558184049]
  We show how memory bandwidth can become the dominant bottleneck for decoder models. We argue for a redesign in model architecture, training, and deployment strategies to overcome this memory limitation.
  arXiv  Detail & Related papers  (2024-03-21T04:31:59Z)
• JORA: JAX Tensor-Parallel LoRA Library for Retrieval Augmented Fine-Tuning [16.86356520836045]
  We introduce a novel framework for PEFT-compatible fine-tuning of Llama-2 models, leveraging distributed training. Our framework uniquely utilizes JAX's just-in-time (JIT) compilation and tensor-sharding for efficient resource management. Our experiments show more than 12x improvement in runtime compared to Hugging Face/DeepSpeed implementation with four GPU while consuming less than half the VRAM per GPU.
  arXiv  Detail & Related papers  (2024-03-17T23:02:04Z)
• FusionAI: Decentralized Training and Deploying LLMs with Massive Consumer-Level GPUs [57.12856172329322]
  We envision a decentralized system unlocking the potential vast untapped consumer-level GPU. This system faces critical challenges, including limited CPU and GPU memory, low network bandwidth, the variability of peer and device heterogeneity.
  arXiv  Detail & Related papers  (2023-09-03T13:27:56Z)
• DistTGL: Distributed Memory-Based Temporal Graph Neural Network Training [18.52206409432894]
  DistTGL is an efficient and scalable solution to train memory-based TGNNs on distributed GPU clusters. In experiments, DistTGL achieves near-linear convergence speedup, outperforming state-of-the-art single-machine method by 14.5% in accuracy and 10.17x in training throughput.
  arXiv  Detail & Related papers  (2023-07-14T22:52:27Z)
• Accelerating Sampling and Aggregation Operations in GNN Frameworks with GPU Initiated Direct Storage Accesses [9.773813896475264]
  Graph Neural Networks (GNNs) are emerging as a powerful tool for learning from graph-structured data. Training GNNs on large-scale graphs remains a significant challenge due to lack of efficient data access and data movement methods. We propose the GPU Initiated Direct Storage Access (GIDS) dataloader to enable GPU-oriented GNN training for large-scale graphs.
  arXiv  Detail & Related papers  (2023-06-28T17:22:15Z)
• ElegantRL-Podracer: Scalable and Elastic Library for Cloud-Native Deep Reinforcement Learning [141.58588761593955]
  We present a library ElegantRL-podracer for cloud-native deep reinforcement learning. It efficiently supports millions of cores to carry out massively parallel training at multiple levels. At a low-level, each pod simulates agent-environment interactions in parallel by fully utilizing nearly 7,000 GPU cores in a single GPU.
  arXiv  Detail & Related papers  (2021-12-11T06:31:21Z)
• PatrickStar: Parallel Training of Pre-trained Models via a Chunk-based Memory Management [19.341284825473558]
  Pre-trained model (PTM) is revolutionizing Artificial intelligence (AI) technology. PTM learns a model with general language features on the vast text and then fine-tunes the model using a task-specific dataset. PatrickStar reduces memory requirements of computing platforms by using heterogeneous memory space.
  arXiv  Detail & Related papers  (2021-08-12T15:58:12Z)
• Improving Computational Efficiency in Visual Reinforcement Learning via Stored Embeddings [89.63764845984076]
  We present Stored Embeddings for Efficient Reinforcement Learning (SEER) SEER is a simple modification of existing off-policy deep reinforcement learning methods. We show that SEER does not degrade the performance of RLizable agents while significantly saving computation and memory.
  arXiv  Detail & Related papers  (2021-03-04T08:14:10Z)

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.