Search for: All records

In this paper, we present work towards the development of a new data analytics and machine learning (ML) framework, called MagmaDNN. Our main goal is to provide scalable, high-performance data analytics and ML solutions for scientific applications running on current and upcoming heterogeneous many-core GPU-accelerated architectures. To this end, since many of the functionalities needed are based on standard linear algebra (LA) routines, we designed MagmaDNN to derive its performance power from the MAGMA library. The close integration provides the fundamental (scalable high-performance) LA routines available in MAGMA as a backend to MagmaDNN. We present some design issues for performance and scalability that are specific to ML using Deep Neural Networks (DNN), as well as the MagmaDNN designs towards overcoming them. In particular, MagmaDNN uses well established HPC techniques from the area of dense LA, including task-based parallelization, DAG representations, scheduling, mixed-precision algorithms, asynchronous solvers, and autotuned hyperparameter optimization. We illustrate these techniques and their incorporation and use to outperform other frameworks, currently available. 

This analysis focuses on a smartphone app known as “Transit” that is used to unlock shared bicycles in Chicago. Data from the app were utilized in a three-part analysis. First, Transit app bikeshare usage patterns were compared with system-wide bikeshare utilization using publicly available data. The results revealed that hourly usage on weekdays generally follows classical peaked commuting patterns; however, daily usage reached its highest level on weekends. This suggests that there may be large numbers of both commuting and recreational users. The second part aimed to identify distinct user groups via cluster analysis; the results revealed six different clusters: (1) commuters, (2) utility users, (3) leisure users, (4) infrequent commuters, (5) weekday visitors, and (6) weekend visitors. The group unlocking the most shared bikes (45.58% of all Transit app unlocks) was commuters, who represent 10% of Transit app bikeshare users. The third part proposed a trip chaining algorithm to identify “trip chaining bikers.” This term refers to bikeshare users who return a shared bicycle and immediately check out another, presumably to avoid paying extra usage fees for trips over 30 min. The algorithm revealed that 27.3% of Transit app bikeshare users exhibited this type of “bike chaining” behavior. However, this varied substantially between user groups; notably, 66% of Transit app bikeshare users identified as commuters made one or more bike chaining unlocks. The implications are important for bikeshare providers to understand the impact of pricing policies, particularly in encouraging the turn-over of bicycles. 

MagmaDNN [17] is a deep learning framework driven using the highly optimized MAGMA dense linear algebra package. The library offers comparable performance to other popular frameworks, such as TensorFlow, PyTorch, and Theano. C++ is used to implement the framework providing fast memory operations, direct cuda access, and compile time errors. Common neural network layers such as Fully Connected, Convolutional, Pooling, Flatten, and Dropout are included. Hyperparameter tuning is performed with a parallel grid search engine. MagmaDNN uses several techniques to accelerate network training. For instance, convolutions are performed using the Winograd algorithm and FFTs. Other techniques include MagmaDNNs custom memory manager, which is used to reduce expensive memory transfers, and accelerated training by distributing batches across GPU nodes. This paper provides an overview of the MagmaDNN framework and how it leverages the MAGMA library to attain speed increases. This paper also addresses how deep networks are accelerated by training in parallel and further challenges with parallelization. 

openDIEL is a workflow engine that aims to give researchers and users of HPC an efficient way to coordinate, organize, and interconnect many disparate modules of computation in order to effectively utilize and allocate HPC resources [13]. A GUI has been developed to aid in creating workflows, and allows for the specification of data science jobs, including specification neural network architectures, data processing, and hyperparameter tuning. Existing machine learning tools can be readily used in the openDIEL, allowing for easy experimentation with various models and approaches. 

The Fast Fourier Transform is a fundamental tool in scientific and technical computation. The highly parallelizable nature of the algorithm makes it a suitable candidate for GPU acceleration. This paper focuses on exploiting the speedup due to using the half precision multiplication capability of the latest GPUs' tensor core hardware without significantly degrading the precision of the Fourier Transform result. We develop an algorithm that dynamically splits the input single precision dataset into two half precision sets at the lowest level, uses half precision multiplication, and recombines the result at a later step. This work paves the way for using tensor cores for high precision inputs.