Engineering Papers⌕ Search

Engineering topics

Emani, Murali

Publications and source records attributed to Emani, Murali.

18 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

MoE-Inference-Bench: Performance Evaluation of Mixture of Expert Large Language and Vision Models

Mixture of Experts (MoE) models have enabled the scaling of Large Language Models (LLMs) and Vision Language Models (VLMs) by achieving massive parameter counts while maintaining computational efficiency. However, MoEs introduce several inference-time challenges, including load imbalance across experts and the additional routing computational overhead. To address these challenges and fully harness the benefits of MoE, a systematic evaluation of hardware acceleration techniques is essential. We present MoE-Inference-Bench, a comprehensive study to evaluate MoE performance across diverse scenarios. We analyze the impact of batch size, sequence length, and critical MoE hyperparameters such as FFN dimensions and number of experts on throughput. We evaluate several optimization techniques on Nvidia H100 GPUs, including pruning, Fused MoE operations, speculative decoding, quantization, and various parallelization strategies. Our evaluation includes MoEs from the Mixtral, DeepSeek, OLMoE and Qwen families. The results reveal performance differences across configurations and provide insights for the efficient deployment of MoEs.

Chitty-Venkata, Krishna Teja↗
DOE OSTI2025

LLM-Inference-Bench: Inference Benchmarking of Large Language Models on AI Accelerators

Large Language Models (LLMs) have propelled groundbreaking advancements across several domains and are commonly used for text generation applications. However, the computational demands of these complex models pose significant challenges, requiring efficient hardware acceleration. Benchmarking the performance of LLMs across diverse hardware platforms is crucial to understanding their scalability and throughput characteristics. We introduce LLM-Inference-Bench, a comprehensive benchmarking suite to evaluate the hardware inference performance of LLMs. We thoroughly analyze diverse hardware platforms, including GPUs from Nvidia and AMD and specialized AI accelerators, Intel Habana and SambaNova. Our evaluation includes several LLM inference frameworks and models from LLaMA, Mistral, and Qwen families with 7B and 70B parameters. Our benchmarking results reveal the strengths and limitations of various models, hardware platforms, and inference frameworks. We provide an interactive dashboard to help identify configurations for optimal performance for a given hardware platform.

Chitty-Venkata, Krishna Teja↗
DOE OSTI2024

HPC-FAIR: A Framework Managing Data and AI Models for Analyzing and Optimizing Scientific Applications

The increasing reliance on machine learning (ML) to analyze and optimize large-scale scientific applications on supercomputers faces a significant bottleneck: the lack of readily available, high-quality training datasets and the difficulty in reusing existing AI models. This project was motivated by the urgent need to address the “FAIR” principles (Findability, Accessibility, Interoperability, Reusability) for both training datasets and AI models in the high-performance computing (HPC) domain. The project developed HPC-FAIR, a high-performance computing data management framework designed to centralize HPC-related datasets and AI models within a unified hub. To ensure interoperability, the framework established a standardized representation and vocabulary (ontology) for both data and models. HPC-FAIR also implemented automated workflows to streamline data processing, model access, and benchmarking. Additionally, the project focused on optimizing data harnessing efficiency through advanced techniques like deep reuse and compression-based analytics.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Cross-Feature Transfer Learning for Efficient Tensor Program Generation

Tuning tensor program generation involves navigating a vast search space to find optimal program transformations and measurements for a program on the target hardware. The complexity of this process is further amplified by the exponential combinations of transformations, especially in heterogeneous environments. This research addresses these challenges by introducing a novel approach that learns the joint neural network and hardware features space, facilitating knowledge transfer to new, unseen target hardware. A comprehensive analysis is conducted on the existing state-of-the-art dataset, TenSet, including a thorough examination of test split strategies and the proposal of methodologies for dataset pruning. Leveraging an attention-inspired technique, we tailor the tuning of tensor programs to embed both neural network and hardware-specific features. Notably, our approach substantially reduces the dataset size by up to 53% compared to the baseline without compromising Pairwise Comparison Accuracy (PCA). Furthermore, our proposed methodology demonstrates competitive or improved mean inference times with only 25–40% of the baseline tuning time across various networks and target hardware. The attention-based tuner can effectively utilize schedules learned from previous hardware program measurements to optimize tensor program tuning on previously unseen hardware, achieving a top-5 accuracy exceeding 90%. This research introduces a significant advancement in autotuning tensor program generation, addressing the complexities associated with heterogeneous environments and showcasing promising results regarding efficiency and accuracy.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Toward a Holistic Performance Evaluation of Large Language Models Across Diverse AI Accelerators

Artificial intelligence (AI) methods have become critical in scientific applications to help accelerate scientific discovery. Large language models (LLMs) are being considered a promising approach to address some challenging problems because of their superior generalization capabilities across domains. The effectiveness of the models and the accuracy of the applications are contingent upon their efficient execution on the underlying hardware infrastructure. Specialized Al accelerator hardware systems have recently become available for accelerating Al applications. However, the comparative performance of these AI accelerators on large language models has not been previously studied. In this paper, we systematically study LLMs on multiple AI accelerators and GPUs and evaluate their performance characteristics for these models. We evaluate these systems with (i) a micro-benchmark using a core transformer block, (ii) a GPT-2 model, and (iii) an 1,I,M-driven science use case, GenSLM. We present our findings and analyses of the models' performance to better understand the intrinsic capabilities of AI accelerators. Furthermore, our analysis takes into account key factors such as sequence lengths, scaling behavior, and sensitivity to gradient accumulation steps.

Emani, Murali↗
DOE OSTI2023

GenSLMs: Genome-scale language models reveal SARS-CoV-2 evolutionary dynamics

We seek to transform how new and emergent variants of pandemic-causing viruses, specifically SARS-CoV-2, are identified and classified. By adapting large language models (LLMs) for genomic data, we build genome-scale language models (GenSLMs) which can learn the evolutionary landscape of SARS-CoV-2 genomes. By pre-training on over 110 million prokaryotic gene sequences and fine-tuning a SARS-CoV-2-specific model on 1.5 million genomes, we show that GenSLMs can accurately and rapidly identify variants of concern. Thus, to our knowledge, GenSLMs represents one of the first whole-genome scale foundation models which can generalize to other prediction tasks. We demonstrate scaling of GenSLMs on GPU-based supercomputers and AI-hardware accelerators utilizing 1.63 Zettaflops in training runs with a sustained performance of 121 PFLOPS in mixed precision and peak of 850 PFLOPS. We present initial scientific insights from examining GenSLMs in tracking evolutionary dynamics of SARS-CoV-2, paving the path to realizing this on large biological data.

Zvyagin, Maxim↗
DOE OSTI2023

Differentiable Neural Architecture, Mixed Precision and Accelerator Co-Search

Quantization, effective Neural Network architecture, and efficient accelerator hardware are three important design paradigms to maximize accuracy and efficiency. Mixed Precision Quantization is a process of assigning different precision to different Neural Network layers for optimized inference. Neural Architecture Search (NAS) is a process of automatically designing the neural network for a task and can also be extended to search for the precision of each weight and activation matrix. In this paper, we develop the following three methods: (i) Fast Differentiable Hardware-aware Mixed Precision Quantization Search method to find optimal precision, (ii) Joint Differentiable hardware-aware Architecture and Mixed Precision Quantization Co-search, (iii) Joint Accelerator, Architecture, and Precision triple co-search to find best possibilities in all the three worlds. We demonstrate the effectiveness of our proposed methods targeting Bitfusion accelerator by searching mixed precision models on MobilenetV2. We achieve better accuracy-latency trade-off models than the manually designed and previously proposed search methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A survey of techniques for optimizing transformer inference

Recent years have seen a phenomenal rise in the performance and applications of transformer neural networks. The family of transformer networks, including Bidirectional Encoder Representations from Transformer (BERT), Generative Pretrained Transformer (GPT) and Vision Transformer (ViT), have shown their effectiveness across Natural Language Processing (NLP) and Computer Vision (CV) domains. Transformer-based networks such as ChatGPT have impacted the lives of common men. However, the quest for high predictive performance has led to an exponential increase in transformers' memory and compute footprint. Researchers have proposed techniques to optimize transformer inference at all levels of abstraction. Further, this paper presents a comprehensive survey of techniques for optimizing the inference phase of transformer networks. We survey techniques such as knowledge distillation, pruning, quantization, neural architecture search and lightweight network design at the algorithmic level. We further review hardware-level optimization techniques and the design of novel hardware accelerators for transformers. We summarize the quantitative results on the number of parameters/FLOPs and the accuracy of several models/techniques to showcase the tradeoff exercised by them. We also outline future directions in this rapidly evolving field of research. We believe that this survey will educate both novice and seasoned researchers and also spark a plethora of research efforts in this field.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

FAIR for AI: An interdisciplinary and international community building perspective

A foundational set of findable, accessible, interoperable, and reusable (FAIR) principles were proposed in 2016 as prerequisites for proper data management and stewardship, with the goal of enabling the reusability of scholarly data. The principles were also meant to apply to other digital assets, at a high level, and over time, the FAIR guiding principles have been re-interpreted or extended to include the software, tools, algorithms, and workflows that produce data. FAIR principles are now being adapted in the context of AI models and datasets. Here, we present the perspectives, vision, and experiences of researchers from different countries, disciplines, and backgrounds who are leading the definition and adoption of FAIR principles in their communities of practice, and discuss outcomes that may result from pursuing and incentivizing FAIR AI research. The material for this report builds on the FAIR for AI Workshop held at Argonne National Laboratory on June 7, 2022.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Neural Architecture Search Benchmarks: Insights and Survey

Neural Architecture Search (NAS), a promising and fast-moving research field, aims to automate the architectural design of Deep Neural Networks (DNNs) to achieve better performance on the given task and dataset. NAS methods have been very successful in discovering efficient models for various Computer Vision, Natural Language Processing, etc. The major obstacles to the advancement of NAS techniques are the demand for large computation resources and fair evaluation of various search methods. The differences in training pipeline and setting make it challenging to compare the efficiency of two NAS algorithms. A large number of NAS Benchmarks to simulate the architecture evaluation in seconds have been released over the last few years to ease the computation burden of training neural networks and can aid in the unbiased assessment of different search methods. This paper provides an extensive review of several publicly available NAS Benchmarks in the literature. We provide technical details and a deeper understanding of each benchmark and point out future directions.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

2022 AI Testbed Expeditions Report

By exploiting the coherent properties of a light source, coherent diffraction imaging (CDI) is able to obtain the sample image at a nanoscale resolution using the measured diffraction pattern. Bragg Coherent Diffraction Imaging (BCDI) has become valuable for recovering the displacement and strain field of crystals, providing a valuable tool in material science and solid-state physics. X-ray ptychography is another emerging CDI technique that can produce a high-resolution image of the extended sample and has become popular in many research areas (e.g., materials science, biology, electronics, and optics characterization). CDI including BCDI and ptychography has become an established technique in Synchrotron Facilities including the Advanced Photon Source (APS) and will greatly benefit from the 100x coherent flux increase of the upcoming APS Upgrade (APSU). The current image formation process in CDI employs iterative phase retrieval algorithms, which is a time-consuming and computationally expensive process. Especially after APSU, the traditional iterative methods will not be able to match the experimental data acquisition speed. We employ deep learning (DL) approach to replace the iterative approaches, therefore allowing hundreds of times faster recovery of the object. We developed AutoPhaseNN, a DL-based approach which learns to solve the inverse problem without labeled data. Taking 3D BCDI as a representative technique, AutoPhaseNN has been demonstrated to be one hundred times faster than traditional iterative phase retrieval methods while providing comparable image quality. The current network is trained with 64 x 64 x 64 data size, to achieve higher resolution imaging, we will need to scale the network to input and train/infer 3D arrays of size 256 x 256 x 256 (today) and of size 2560x2560x2560 (APSU). However, the scalability of the network is restricted due to the memory-intensive training process. To perform the training for a 256 x 256 x 256 data size, the required memory exceeds the capacity of the current machine. In this project, we explore using Sambanova system to train the network for the direct data inversion for CDI.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Intelligent resolution: Integrating Cryo-EM with AI-driven multi-resolution simulations to observe the severe acute respiratory syndrome coronavirus-2 replication-transcription machinery in action

The severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) replication transcription complex (RTC) is a multi-domain protein responsible for replicating and transcribing the viral mRNA inside a human cell. Attacking RTC function with pharmaceutical compounds is a pathway to treating COVID-19. Conventional tools, e.g., cryo-electron microscopy and all-atom molecular dynamics (AAMD), do not provide sufficiently high resolution or timescale to capture important dynamics of this molecular machine. Consequently, we develop an innovative workflow that bridges the gap between these resolutions, using mesoscale fluctuating finite element analysis (FFEA) continuum simulations and a hierarchy of AI-methods that continually learn and infer features for maintaining consistency between AAMD and FFEA simulations. We leverage a multi-site distributed workflow manager to orchestrate AI, FFEA, and AAMD jobs, providing optimal resource utilization across HPC centers. Our study provides unprecedented access to study the SARS-CoV-2 RTC machinery, while providing general capability for AI-enabled multi-resolution simulations at scale.

Trifan, Anda↗
DOE OSTI2022

Throughput-Oriented and Accuracy-Aware DNN Training with BFloat16 on GPU

Deep Neural Networks (DNNs) have transformed the field of artificial intelligence and achieved extraordinary success in many areas. The training of DNNs is commonly compute and memory-intensive, which has resulted in several optimizations in the training phase. Among them, reduced precision is a typical and widely used technique to accelerate DNN training and reduce memory requirements. However, applying a widely adopted reduced precision format such as Float16 to all involved operations in DNN training is not optimal as the use of Float16 in some operations can hurt model accuracy. Meanwhile, additional optimizations including loss scaling and autocast techniques can mitigate the accuracy loss but lead to inherent overhead and inadequate use of reduced precision. In this work, we leverage another reduced precision format, BFloat16, and introduce a throughput-oriented and accuracy-aware approach to maximize the performance potential of DNN training. Since the high throughput provided by BFloat16 format is accompanied by low precision of the floating-point representation, this approach achieves high throughput by using BFloat16 on all DNN operations and avoids the accuracy loss through a customized accuracy-aware normalization. Results show that our approach outperforms the state-of-the-art mixed-precision training by 1.21x on an NVIDIA A100 GPU.

Xie, Zhen↗
DOE OSTI2021

Accelerating Scientific Applications With SambaNova Reconfigurable Dataflow Architecture

Here, our exploratory work finds that the SambaNova Reconfigurable Dataflow Architecture (RDA) along with the SambaFlow software stack provides for an attractive system and solution to accelerate AI for science workloads. We have observed the efficacy of using the system with a diverse set of science applications and reasoned their suitability for performance gains over traditional hardware. As the Data-Scale system provides for a very large memory capacity, the system can be used to train models that typically do not fit in a GPU. The architecture also provides for deeper integration with upcoming supercomputers at the Argonne Leadership Computing Facility (ALCF), a US Department of Energy Office of Science user facility, to help advance science insights.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

EReinit: Scalable and efficient fault-tolerance for bulk-synchronous MPI applications

Scientists from many different fields have been developing Bulk-Synchronous MPI applications to simulate and study a wide variety of scientific phenomena. Since failure rates are expected to increase in larger-scale future HPC systems, providing efficient fault-tolerance mechanisms for this class of applications is paramount. The global-restart model has been proposed to decrease the time of failure recovery in Bulk-Synchronous applications by allowing a fast reinitialization of MPI. However, the current implementations of this model have several drawbacks: they lack efficiency; their scalability have not been shown; and they require the use of the MPI profiling interface, which precludes the use of tools. Here, we present EReinit, an implementation of the global-restart model that addresses these problems. Our key idea and optimization is the co-design of basic fault-tolerance mechanisms such as failure detection, notification, and recovery between MPI and the resource manager in contrast to current approaches on which these mechanisms are implemented in MPI only. We demonstrate EReinit in three HPC programs and show that it is up to four times more efficient than existing solutions at 4,096 processes.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

MPI Stages: Checkpointing MPI State for Bulk Synchronous Applications

When an MPI program experiences a failure, the most common recovery approach is to restart all processes from a previous checkpoint and to re-queue the entire job. A disadvantage of this method is that, although the failure occurred within the main application loop, live processes must start again from the beginning of the program, along with new replacement processes---this incurs unnecessary overhead for live processes. To avoid such overheads and concomitant delays, we introduce the concept of "MPI Stages." MPI Stages saves internal MPI state in a separate checkpoint in conjunction with application state. Upon failure, both MPI and application state are recovered, respectively, from their last synchronous checkpoints and continue without restarting the overall MPI job. Live processes roll back only a few iterations within the main loop instead of rolling back to the beginning of the program, while a replacement of failed process restarts and reintegrates, thereby achieving faster failure recovery. This approach integrates well with large-scale, bulk synchronous applications and checkpoint/restart. In this paper, we identify requirements for production MPI implementations to support state checkpointing with MPI Stages, which includes capturing and managing internal MPI state and serializing and deserializing user handles to MPI objects. We evaluate our fault tolerance approach with a proof-of-concept prototype MPI implementation that includes MPI Stages. We demonstrate its functionality and performance using LULESH and microbenchmarks. Our results show that MPI Stages reduces the recovery time by 13× for LULESH in comparison to checkpoint/restart.

Sultana, Nawrin↗