An official website of the United States government
Constructing k-nearest neighbor (kNN) graphs is a fundamental component in many machine learning and scientific computing applications. Despite its prevalence, efficiently building all-nearest-neighbor graphs at scale on distributed heterogeneous HPC systems remains challenging, especially for large sparse non-integer datasets. We introduce optimizations for algorithms based on forests of random projection trees. Our novel GPU kernels for batched, within leaf, exact searches achieve 1.18× speedup over sparse reference kernels with less peak memory, and up to 19× speedup over CPU for memory-intensive problems. Our library,PyRKNN, implements distributed randomized projection forests for approximate kNN search. Optimizations to reduce and hide communication overhead allow us to achieve 5× speedup, in per iteration performance, relative to GOFMM (another projection tree, MPI-based kNN library), for a 64M 128d dataset on 1,024 processes. On a single-node we achieve speedup over FAISS-GPU for dense datasets and up to 10× speedup over CPU-only libraries.PyRKNNuniquely supports distributed memory kNN graph construction for both dense and sparse coordinates on CPU and GPU accelerators.
more »
« less
Breaking the computation and communication abstraction barrier in distributed machine learning workloads
Recent trends towards large machine learning models require both training and inference tasks to be distributed. Considering the huge cost of training these models, it is imperative to unlock optimizations in computation and communication to obtain best performance. However, the current logical separation between computation and communication kernels in machine learning frameworks misses optimization opportunities across this barrier. Breaking this abstraction can provide many optimizations to improve the performance of distributed workloads. However, manually applying these optimizations requires modifying the underlying computation and communication libraries for each scenario, which is both time consuming and error-prone. Therefore, we present CoCoNet, which contains (i) a domain specific language to express a distributed machine learning program in the form of computation and communication operations, (ii) a set of semantics preserving transformations to optimize the program, and (iii) a compiler to generate jointly optimized communication and computation GPU kernels. Providing both computation and communication as first class constructs allows users to work on a high-level abstraction and apply powerful optimizations, such as fusion or overlapping of communication and computation. CoCoNet enabled us to optimize data-, model- and pipeline-parallel workloads in large language models with only a few lines of code. Our experiments show that CoCoNet significantly outperforms state-of-the-art distributed machine learning implementations.
more »
« less
- Award ID(s):
- 2052696
- PAR ID:
- 10319730
- Date Published:
- Journal Name:
- ASPLOS 2022: Proceedings of the 27th ACM International Conference on Architectural Support for Programming Languages and Operating Systems
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Poole, Steve; Hernandez, Oscar; Baker, Matthew; Curtis, Tony (Ed.)SHMEM-ML is a domain specific library for distributed array computations and machine learning model training & inference. Like other projects at the intersection of machine learning and HPC (e.g. dask, Arkouda, Legate Numpy), SHMEM-ML aims to leverage the performance of the HPC software stack to accelerate machine learning workflows. However, it differs in a number of ways. First, SHMEM-ML targets the full machine learning workflow, not just model training. It supports a general purpose nd-array abstraction commonly used in Python machine learning applications, and efficiently distributes transformation and manipulation of this ndarray across the full system. Second, SHMEM-ML uses OpenSHMEM as its underlying communication layer, enabling high performance networking across hundreds or thousands of distributed processes. While most past work in high performance machine learning has leveraged HPC message passing communication models as a way to efficiently exchange model gradient updates, SHMEM-ML’s focus on the full machine learning lifecycle means that a more flexible and adaptable communication model is needed to support both fine and coarse grain communication. Third, SHMEM-ML works to interoperate with the broader Python machine learning software ecosystem. While some frameworks aim to rebuild that ecosystem from scratch on top of the HPC software stack, SHMEM-ML is built on top of Apache Arrow, an in-memory standard for data formatting and data exchange between libraries. This enables SHMEM-ML to share data with other libraries without creating copies of data. This paper describes the design, implementation, and evaluation of SHMEM-ML – demonstrating a general purpose system for data transformation and manipulation while achieving up to a 38× speedup in distributed training performance relative to the industry standard Horovod framework without a regression in model metrics.more » « less
-
null (Ed.)Deep learning has been shown as a successful method for various tasks, and its popularity results in numerous open-source deep learning software tools. Deep learning has been applied to a broad spectrum of scientific domains such as cosmology, particle physics, computer vision, fusion, and astrophysics. Scientists have performed a great deal of work to optimize the computational performance of deep learning frameworks. However, the same cannot be said for I/O performance. As deep learning algorithms rely on big-data volume and variety to effectively train neural networks accurately, I/O is a significant bottleneck on large-scale distributed deep learning training. This study aims to provide a detailed investigation of the I/O behavior of various scientific deep learning workloads running on the Theta supercomputer at Argonne Leadership Computing Facility. In this paper, we present DLIO, a novel representative benchmark suite built based on the I/O profiling of the selected workloads. DLIO can be utilized to accurately emulate the I/O behavior of modern scientific deep learning applications. Using DLIO, application developers and system software solution architects can identify potential I/O bottlenecks in their applications and guide optimizations to boost the I/O performance leading to lower training times by up to 6.7x.more » « less
-
Scientific communities are increasingly adopting machine learning and deep learning models in their applications to accelerate scientific insights. High performance computing systems are pushing the frontiers of performance with a rich diversity of hardware resources and massive scale-out capabilities. There is a critical need to understand fair and effective benchmarking of machine learning applications that are representative of real-world scientific use cases. MLPerf ™ is a community-driven standard to benchmark machine learning workloads, focusing on end-to-end performance metrics. In this paper, we introduce MLPerf HPC, a benchmark suite of large-scale scientific machine learning training applications, driven by the MLCommons ™ Association. We present the results from the first submission round including a diverse set of some of the world’s largest HPC systems. We develop a systematic framework for their joint analysis and compare them in terms of data staging, algorithmic convergence and compute performance. As a result, we gain a quantitative understanding of optimizations on different subsystems such as staging and on-node loading of data, compute-unit utilization and communication scheduling enabling overall >10× (end-to-end) performance improvements through system scaling. Notably, our analysis shows a scale-dependent interplay between the dataset size, a system’s memory hierarchy and training convergence that underlines the importance of near-compute storage. To overcome the data-parallel scalability challenge at large batch-sizes, we discuss specific learning techniques and hybrid data-and-model parallelism that are effective on large systems. We conclude by characterizing each benchmark with respect to low-level memory, I/O and network behaviour to parameterize extended roofline performance models in future rounds.more » « less
-
Federated Learning (FL) revolutionizes collaborative machine learning among Internet of Things (IoT) devices by enabling them to train models collectively while preserving data privacy. FL algorithms fall into two primary categories: synchronous and asynchronous. While synchronous FL efficiently handles straggler devices, its convergence speed and model accuracy can be compromised. In contrast, asynchronous FL allows all devices to participate but incurs high communication overhead and potential model staleness. To overcome these limitations, the paper introduces a semi-synchronous FL framework that uses client tiering based on computing and communication latencies. Clients in different tiers upload their local models at distinct frequencies, striking a balance between straggler mitigation and communication costs. Building on this, the paper proposes the Dynamic client clustering, bandwidth allocation, and local training for semi-synchronous Federated learning (DecantFed) algorithm to dynamically optimize client clustering, bandwidth allocation, and local training workloads in order to maximize data sample processing rates in FL. DecantFed dynamically optimizes client clustering, bandwidth allocation, and local training workloads for maximizing data processing rates in FL. It also adapts client learning rates according to their tiers, thus addressing the model staleness issue. Extensive simulations using benchmark datasets like MNIST and CIFAR-10, under both IID and non-IID scenarios, demonstrate DecantFed’s superior performance. It outperforms FedAvg and FedProx in convergence speed and delivers at least a 28% improvement in model accuracy, compared to FedProx.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1145/3503222.3507778
