skip to main content

Title: Neocortex and Bridges-2: A High Performance AI+HPC Ecosystem for Science, Discovery, and Societal Good
Artificial intelligence (AI) is transforming research through analysis of massive datasets and accelerating simulations by factors of up to a billion. Such acceleration eclipses the speedups that were made possible though improvements in CPU process and design and other kinds of algorithmic advances. It sets the stage for a new era of discovery in which previously intractable challenges will become surmountable, with applications in fields such as discovering the causes of cancer and rare diseases, developing effective, affordable drugs, improving food sustainability, developing detailed understanding of environmental factors to support protection of biodiversity, and developing alternative energy sources as a step toward reversing climate change. To succeed, the research community requires a high-performance computational ecosystem that seamlessly and efficiently brings together scalable AI, general-purpose computing, and large-scale data management. The authors, at the Pittsburgh Supercomputing Center (PSC), launched a second-generation computational ecosystem to enable AI-enabled research, bringing together carefully designed systems and groundbreaking technologies to provide at no cost a uniquely capable platform to the research community. It consists of two major systems: Neocortex and Bridges-2. Neocortex embodies a revolutionary processor architecture to vastly shorten the time required for deep learning training, foster greater integration of artificial deep learning with more » scientific workflows, and accelerate graph analytics. Bridges-2 integrates additional scalable AI, high-performance computing (HPC), and high-performance parallel file systems for simulation, data pre- and post-processing, visualization, and Big Data as a Service. Neocortex and Bridges-2 are integrated to form a tightly coupled and highly flexible ecosystem for AI- and data-driven research. « less
Nesmachnow, S.; Castro, H.; Tchernykh, A.
Award ID(s):
Publication Date:
Journal Name:
Communications in computer and information science
Page Range or eLocation-ID:
Sponsoring Org:
National Science Foundation
More Like this
  1. Today’s landscape of computational science is evolving rapidly, with a need for new, flexible, and responsive supercomputing platforms for addressing the growing areas of artificial intelligence (AI), data analytics (DA) and convergent collaborative research. To support this community, we designed and deployed the Bridges-2 platform. Building on our highly successful Bridges supercomputer, which was a high-performance computing resource supporting new communities and complex workflows, Bridges-2 supports traditional and nontraditional research communities and applications; integrates new technologies for converged, scalable high-performance computing (HPC), AI, and data analytics; prioritizes researcher productivity and ease of use; and provides an extensible architecture for interoperation with complementary data intensive projects, campuses, and clouds. In this report, we describe Bridges-2’s hardware and configuration, user environments, and systems support and present the results of the successful Early User Program.
  2. To advance knowledge by enabling unprecedented AI speed and scalability, the Pittsburgh Supercomputing Center (PSC), a joint research center of Carnegie Mellon University and the University of Pittsburgh, in partnership with Cerebras Systems and Hewlett Packard Enterprise (HPE), has deployed Neocortex, an innovative computing platform that accelerates scientific discovery by vastly shortening the time required for deep learning training and inference, fosters greater integration of deep AI models with scientific workflows, and provides promising hardware for the development of more efficient algorithms for artificial intelligence and graph analytics. Neocortex advances knowledge by accelerating scientific research, enabling development of more accurate models and use of larger training data, scaling model parallelism to unprecedented levels, and focusing on human productivity by simplifying tuning and hyperparameter optimization to create a transformative hardware and software platform for the exploration of new frontiers. Neocortex has been integrated with PSC’s complementary infrastructure. This papers shares experiences, decisions, and findings made in that process. The system is serving science and engineering users via an early user access program. Valuable artifacts developed during the integration phase have been made available via a public repository and have been consulted by other AI system deployments that have seen Neocortex asmore »an inspiration.« less
  3. The National Ecological Observatory Network (NEON) is a continental-scale observatory with sites across the US collecting standardized ecological observations that will operate for multiple decades. To maximize the utility of NEON data, we envision edge computing systems that gather, calibrate, aggregate, and ingest measurements in an integrated fashion. Edge systems will employ machine learning methods to cross-calibrate, gap-fill and provision data in near-real time to the NEON Data Portal and to High Performance Computing (HPC) systems, running ensembles of Earth system models (ESMs) that assimilate the data. For the first time gridded EC data products and response functions promise to offset pervasive observational biases through evaluating, benchmarking, optimizing parameters, and training new ma- chine learning parameterizations within ESMs all at the same model-grid scale. Leveraging open-source software for EC data analysis, we are al- ready building software infrastructure for integration of near-real time data streams into the International Land Model Benchmarking (ILAMB) package for use by the wider research community. We will present a perspective on the design and integration of end-to-end infrastructure for data acquisition, edge computing, HPC simulation, analysis, and validation, where Artificial Intelligence (AI) approaches are used throughout the distributed workflow to improve accuracy and computational performance.
  4. Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)
    The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should producemore »identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues.« less
  5. e. With recent advances in online sensing technology and high-performance computing, structural health monitoring (SHM) has begun to emerge as an automated approach to the real-time conditional monitoring of civil infrastructure. Ideal SHM strategies detect and characterize damage by leveraging measured response data to update physics-based finite element models (FEMs). When monitoring composite structures, such as reinforced concrete (RC) bridges, the reliability of FEM based SHM is adversely affected by material, boundary, geometric, and other model uncertainties. Civil engineering researchers have adapted popular artificial intelligence (AI) techniques to overcome these limitations, as AI has an innate ability to solve complex and ill-defined problems by leveraging advanced machine learning techniques to rapidly analyze experimental data. In this vein, this study employs a novel Bayesian estimation technique to update a coupled vehicle-bridge FEM for the purposes of SHM. Unlike existing AI based techniques, the proposed approach makes intelligent use of an embedded FEM model, thus reducing the parameter space while simultaneously guiding the Bayesian model via physics-based principles. To validate the method, bridge response data is generated from the vehicle-bridge FEM given a set of “true” parameters and the bias and standard deviation of the parameter estimates are analyzed. Additionally, the meanmore »parameter estimates are used to solve the FEM model and the results are compared against the results obtained for “true” parameter values. A sensitivity study is also conducted to demonstrate methods for properly formulating model spaces to improve the Bayesian estimation routine. The study concludes with a discussion highlighting factors that need to be considered when leveraging experimental data to update FEMs of concrete structures using AI techniques.« less