skip to main content

Explore Research Products in the NSF-PAR

Title: Memory-Efficient Pipeline-Parallel DNN Training
Many state-of-the-art ML results have been obtained by scaling up the number of parameters in existing models. However, parameters and activations for such large models often do not fit in the memory of a single accelerator device; this means that it is necessary to distribute training of large models over multiple accelerators. In this work, we propose PipeDream-2BW, a system that supports memory-efficient pipeline parallelism. PipeDream-2BW uses a novel pipelining and weight gradient coalescing strategy, combined with the double buffering of weights, to ensure high throughput, low memory footprint, and weight update semantics similar to data parallelism. In addition, PipeDream-2BW automatically partitions the model over the available hardware resources, while respecting hardware constraints such as memory capacities of accelerators and interconnect topologies. PipeDream-2BW can accelerate the training of large GPT and BERT language models by up to 20x with similar final model accuracy.  more » « less
Award ID(s):
1651570
NSF-PAR ID:
10327318
Author(s) / Creator(s):
; ; ; ;
Date Published:
Journal Name:
Proceedings of Machine Learning Research
Volume:
139
ISSN:
2640-3498
Page Range / eLocation ID:
7937-7947
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ( , 53rd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO))
    null (Ed.)
    Graph processing workloads are memory intensive with irregular access patterns and large memory footprint resulting in low data locality. Their popular software implementations typically employ either Push or Pull style propagation of changes through the graph over multiple iterations that follow the Bulk Synchronous Model. The performance of these algorithms on traditional computing systems is limited by random reads/writes of vertex values, synchronization overheads, and additional overheads for tracking active sets of vertices or edges across iterations. In this paper, we present GraphPulse, a hardware framework for asynchronous graph processing with event-driven scheduling that overcomes the performance limitations of software frameworks. Event-driven computation model enables a parallel dataflow-style execution where atomic updates and active sets tracking are inherent to the model; thus, scheduling complexity is reduced and scalability is enhanced. The dataflow nature of the architecture also reduces random reads of vertex values by carrying the values in the events themselves. We capitalize on the update properties commonly present in graph algorithms to coalesce in-flight events and substantially reduce the event storage requirement and the processing overheads incurred. GraphPulse event-model naturally supports asynchronous graph processing, enabling substantially faster convergence by exploiting available parallelism, reducing work, and eliminating synchronization at iteration boundaries. The framework provides easy to use programming interface for faster development of hardware graph accelerators. A single GraphPulse accelerator achieves up to 74x speedup (28x on average) over Ligra, a state of the art software framework, running on a 12 core CPU. It also achieves an average of 6.2x speedup over Graphicionado, a state of the art graph processing accelerator. 
    more » « less
  2. ; ; ; ; ; ; ; ( , Proceedings of the VLDB Endowment)
    Deep neural networks (DNNs) are becoming increasingly deeper, wider, and non-linear due to the growing demands on prediction accuracy and analysis quality. Training wide and deep neural networks require large amounts of storage resources such as memory because the intermediate activation data must be saved in the memory during forward propagation and then restored for backward propagation. However, state-of-the-art accelerators such as GPUs are only equipped with very limited memory capacities due to hardware design constraints, which significantly limits the maximum batch size and hence performance speedup when training large-scale DNNs. Traditional memory saving techniques either suffer from performance overhead or are constrained by limited interconnect bandwidth or specific interconnect technology. In this paper, we propose a novel memory-efficient CNN training framework (called COMET) that leverages error-bounded lossy compression to significantly reduce the memory requirement for training in order to allow training larger models or to accelerate training. Our framework purposely adopts error-bounded lossy compression with a strict error-controlling mechanism. Specifically, we perform a theoretical analysis on the compression error propagation from the altered activation data to the gradients, and empirically investigate the impact of altered gradients over the training process. Based on these analyses, we optimize the error-bounded lossy compression and propose an adaptive error-bound control scheme for activation data compression. Experiments demonstrate that our proposed framework can significantly reduce the training memory consumption by up to 13.5X over the baseline training and 1.8X over another state-of-the-art compression-based framework, respectively, with little or no accuracy loss. 
    more » « less
  3. ; ; ; ; ; ; ; ( , SOSP)
    DNN training is extremely time-consuming, necessitating efficient multi-accelerator parallelization. Current approaches to parallelizing training primarily use intra-batch parallelization, where a single iteration of training is split over the available workers, but suffer from diminishing returns at higher worker counts. We present PipeDream, a system that adds inter-batch pipelining to intra-batch parallelism to further improve parallel training throughput, helping to better overlap computation with communication and reduce the amount of communication when possible. Unlike traditional pipelining, DNN training is bi-directional, where a forward pass through the computation graph is followed by a backward pass that uses state and intermediate data computed during the forward pass. Naïve pipelining can thus result in mismatches in state versions used in the forward and backward passes, or excessive pipeline flushes and lower hardware efficiency. To address these challenges, PipeDream versions model parameters for numerically correct gradient computations, and schedules forward and backward passes of different minibatches concurrently on different workers with minimal pipeline stalls. PipeDream also automatically partitions DNN layers among workers to balance work and minimize communication. Extensive experimentation with a range of DNN tasks, models, and hardware configurations shows that PipeDream trains models to high accuracy up to 5.3X faster than commonly used intra-batch parallelism techniques. 
    more » « less
  4. ; ; ; ; ( , IEEE International Symposium on High Performance Computer Architecture (HPCA))
    Graph processing recently received intensive interests in light of a wide range of needs to understand relationships. It is well-known for the poor locality and high memory bandwidth requirement. In conventional architectures, they incur a significant amount of data movements and energy consumption which motivates several hardware graph processing accelerators. The current graph processing accelerators rely on memory access optimizations or placing computation logics close to memory. Distinct from all existing approaches, we leverage an emerging memory technology to accelerate graph processing with analog computation. This paper presents GRAPHR, the first ReRAM-based graph processing accelerator. GRAPHR follows the principle of near-data processing and explores the opportunity of performing massive parallel analog operations with low hardware and energy cost. The analog computation is suitable for graph processing because: 1) The algorithms are iterative and could inherently tolerate the imprecision; 2) Both probability calculation (e.g., PageRank and Collaborative Filtering) and typical graph algorithms involving integers (e.g., BFS/SSSP) are resilient to errors. The key insight of GRAPHR is that if a vertex program of a graph algorithm can be expressed in sparse matrix vector multiplication (SpMV), it can be efficiently performed by ReRAM crossbar. We show that this assumption is generally true for a large set of graph algorithms. GRAPHR is a novel accelerator architecture consisting of two components: memory ReRAM and graph engine (GE). The core graph computations are performed in sparse matrix format in GEs (ReRAM crossbars). The vector/matrix-based graph computation is not new, but ReRAM offers the unique opportunity to realize the massive parallelism with unprecedented energy efficiency and low hardware cost. With small subgraphs processed by GEs, the gain of performing parallel operations overshadows the wastes due to sparsity. The experiment results show that GRAPHR achieves a 16.01X (up to 132.67X) speedup and a 33.82X energy saving on geometric mean compared to a CPU baseline system. Compared to GPU, GRAPHR achieves 1.69X to 2.19X speedup and consumes 4.77X to 8.91X less energy. GRAPHR gains a speedup of 1.16X to 4.12X, and is 3.67X to 10.96X more energy efficiency compared to PIM-based architecture. 
    more » « less
  5. ( , Medical Imaging with Deep Learning (MIDL))
    Deep learning has made great strides in medical imaging, enabled by hardware advances in GPUs. One major constraint for the development of new models has been the saturation of GPU memory resources during training. This is especially true in computational pathology, where images regularly contain more than 1 billion pixels. These pathological images are traditionally divided into small patches to enable deep learning due to hardware limitations. In this work, we explore whether the shared GPU/CPU memory architecture on the M1 Ultra systems-on-a-chip (SoCs) recently released by Apple, Inc. may provide a solution. These affordable systems (less than $5000) provide access to 128 GB of unified memory (Mac Studio with M1 Ultra SoC). As a proof of concept for gigapixel deep learning, we identified tissue from background on gigapixel areas from whole slide images (WSIs). The model was a modified U-Net (4492 parameters) leveraging large kernels and high stride. The M1 Ultra SoC was able to train the model directly on gigapixel images (16000×64000 pixels, 1.024 billion pixels) with a batch size of 1 using over 100 GB of unified memory for the process at an average speed of 1 minute and 21 seconds per batch with Tensorflow 2/Keras. As expected, the model converged with a high Dice score of 0.989 ± 0.005. Training up until this point took 111 hours and 24 minutes over 4940 steps. Other high RAM GPUs like the NVIDIA A100 (largest commercially accessible at 80 GB, ∼$15000) are not yet widely available (in preview for select regions on Amazon Web Services at $40.96/hour as a group of 8). This study is a promising step towards WSI-wise end-to-end deep learning with prevalent network architectures. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email