SEARCH · Search NASA

Results for “privacy”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 109 records · Page 6

All sources NASA NTRS DOE OSTI

DOE OSTI2025

CAFE AU LAIT: Compute-Aware Federated Augmented Low-Rank AI Training

Federated finetuning is crucial for unlocking the knowledge embedded in pretrained Large Language Models (LLMs) when data are geographically distributed across clients. Unlike finetuning with data from a single institution, federated finetuning allows collaboration across multiple institutions, enabling the utilization of diverse and decentralized datasets while preserving data privacy. Given the high computing costs of LLM training and the emphasis on energy efficiency in Federated Learning (FL), Low-Rank Adaptation (LoRA) has emerged as a widely adopted algorithm due to its significantly reduced number of trainable parameters. However, this assumes that all data silos have the necessary computing resources to compute local updates of LLMs. Nevertheless, in practice, the computing resources across clients are highly heterogeneous: while some may have access to hundreds of GPUs, others might have limited or no GPU access. Recently, federated finetuning using synthetic data has been proposed, allowing clients to participate in a collaborative training run without training LLMs locally. However, our experimental results reveal a performance gap between models trained using synthetic data and those trained using local updates. Motivated by the observed heterogeneity in computing resources and the performance gap, we propose a novel two-stage algorithm that leverages the storage and computing capabilities of a strong server. In the first stage, under the coordination of the strong server, clients with limited computing resources collaborate to generate synthetic data, which is transferred to and stored on the strong server. In the second stage, the strong server uses this synthetic data on behalf of the resource-constrained clients to perform federated LoRA finetuning alongside clients with sufficient computing resources. This approach ensures that all clients can participate in the finetuning process. Experimental results demonstrate that incorporating local updates from even a small fraction of clients improves performance compared to using synthetic data for all clients. Furthermore, we incorporate the Gaussian mechanism in both stages to guarantee client-level differential privacy.

Wang, Jiayi [ORNL]↗

DOE OSTI2026

Scalable Federated Learning for Scientific Foundation Models on Leadership-Class Systems

Federated learning (FL) at leadership-class HPC systems remains largely unexplored, despite growing interest in deploying federated workflows on modern HPC systems. This paper provides the first system-level empirical characterization of federated fine-tuning of pretrained foundation models on an exascale supercomputer under a multi-node deployment. Using up to 96 concurrent FL clients deployed across Frontier nodes, we study the impact of client scale, model size, data heterogeneity, partial participation, and differential privacy on runtime, communication overhead, and convergence stability. Our results show that pretrained transformer models remain robust to heterogeneity, client dropout, and privacy noise, while system efficiency degrades rapidly with scale as synchronizat and orchestration dominate runtime. We further demonstrate that system-aware execution strategies, including intra-node aggregation and early aggregation, significantly reduce wall-clock time without degrading model quality. These findings establish a practical performance baseline and inform the design of communication-efficient FL systems on leadership-class HPC platforms.

Kotevska, Olivera [ORNL] (ORCID:0000000316772243)↗

DOE OSTI2024

End-Use Savings Shapes Measure Documentation: Window Film

This documentation focuses on a single end-use savings shape measure - window film. The window film studied in this analysis, called solar control film, is a passive retrofit solution for windows that does not involve window replacement. This type of film is composed of transparent, tinted, or metalized laminated polyester layers and can be attached to an existing window surface (either on the exterior or interior side of the window). The properties of the window film are designed to shift the thermal and optical performances of the overall glazing system in order to serve various needs the customer would have (e.g., heat, glare). While the practical goal of purchasing and installing a window film varies widely in the real market, this study only focuses on the goal of energy savings, and highlighting the corresponding emissions. Other important aspects that customers typically consider include visual comfort, privacy, aesthetics, ultraviolet protection, etc. Thus, in practice, customers often choose a window film product not only to save energy (or cost) but also to mitigate issues around glare, excessive light, daytime privacy, or inconsistent appearance of the building. Window film products that were modeled in this analysis significantly reduced the solar heat gain coefficient of the overall glazing system, resulting in better energy savings for buildings in hot climate regions. However, significantly reducing the solar heat gain coefficient that can block unfavorable heat during the summer can actually harm blocking favorable heat during the winter. By applying window films on a stock of buildings covering various load and weather conditions, this analysis highlights when (e.g., time of day) and where (e.g., geospacial location) we can save energy with window films.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2024

4th Big Data for Nuclear Power Plants Workshop 2023

The Ohio State University and Idaho National Laboratory organized the 4 th Big Data for Nuclear Power Plants Workshop in November, 2023 in Columbus, Ohio. Workshop topics were chosen to understand the challenges and gaps that need to be addressed to maximize the impact of data on the nuclear industry, as well as the associated applications and risks. Discussions were focused around six specific application areas: Operation and Maintenance; Machine Learning in Nuclear Materials and Advanced Manufacturing; Cybersecurity; High-Performance Computing and Massive Computation; Big Data and Digital Twins; and Nuclear Non-Proliferation. The opportunities, challenges, and risks identified in the six focus areas explored in this workshop are diverse, but some common themes emerge, such as the importance of data integrity, quality, coverage, privacy, and traceability. Big data and AI/ML tools can be leveraged to reduce costs, optimize human tasking, and reduce human error across various application areas. In order for the nuclear industry to benefit from big data and advanced analytic capabilities, it is essential to address challenges and risks, such as data privacy, model reliability, and computational resource availability. Learning from other industries that have successfully implemented big data and AI/ML technologies, like the aerospace industry, can help the nuclear industry successfully integrate these technologies.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2025

Scale-up Unlearnable Examples Learning with High-performance Computing

Recent advancements in AI models, like ChatGPT, are structured to retain user interactions, which could inadvertently include sensitive healthcare data. In the healthcare field, particularly when radiologists use AI-driven diagnostic tools hosted on online platforms, there is a risk that medical imaging data may be repurposed for future AI training without explicit consent, spotlighting critical privacy and intellectual property concerns around healthcare data usage. Addressing these privacy challenges, a novel approach known as Unlearnable Examples (UEs) has been introduced, aiming to make data unlearnable to deep learning models. A prominent method within this area, called Unlearnable Clustering (UC), has shown improved UE performance with larger batch sizes but was previously limited by computational resources (e.g., a single workstation). To push the boundaries of UE performance with theoretically unlimited resources, we scaled up UC learning across various datasets using Distributed Data Parallel (DDP) training on the Summit supercomputer. Our goal was to examine UE efficacy at high-performance computing (HPC) levels to prevent unauthorized learning and enhance data security, particularly exploring the impact of batch size on UE’s unlearnability. Utilizing the robust computational capabilities of the Summit, extensive experiments were conducted on diverse datasets such as Pets, MedMNist, Flowers, and Flowers102. Our findings reveal that both overly large and overly small batch sizes can lead to performance instability and affect accuracy. However, the relationship between batch size and unlearnability varied across datasets, highlighting the necessity for tailored batch size strategies to achieve optimal data protection. The use of Summit’s high-performance GPUs, along with the efficiency of the DDP framework, facilitated rapid updates of model parameters and consistent training across nodes. Our results underscore the critical role of selecting appropriate batch sizes based on the specific characteristics of each dataset to prevent learning and ensure data security in deep learning applications. The source code is publicly available at https: // github. com/ hrlblab/ UE_ HPC .

Zhu, Yanfan [Vanderbilt University, Nashville, TN,↗

DOE OSTI2025

Machine Learning for Anomaly Detection in Neural Network Security and SRF Cavities

This dissertation explores the development and deployment of machine learning approaches to address critical challenges in anomaly detection across two distinct domains: neural network security in federated learning settings and cavity behavior analysis in particle accelerator operations at Jefferson Lab in Newport News, Virginia. Anomaly detection identifies deviations from expected patterns, safeguarding systems in cybersecurity, industry, and research against malicious activities and failures. This dissertation demonstrates how our machine learning approaches enhance detection accuracy and efficiency in both neural network security and industrial applications. First, we investigate vulnerabilities in deep neural networks deployed in federated learning. Although federated learning preserves user privacy by training models locally, it remains vulnerable to backdoor attacks, in which malicious participants embed hidden triggers that induce targeted misbehavior. We propose a self-supervised contrastive learning framework to detect and mitigate such backdoor attacks. In our experiments, this method achieves higher detection accuracy and lower false positive rates than existing defenses, while operating without access to local model updates or original training data and thus preserving the privacy guarantees of the federated setting. Second, we address the operational reliability of superconducting radio-frequency (SRF) cavities at the Continuous Electron Beam Accelerator Facility (CEBAF). Our research leverages an unsupervised learning approach, combined with Principal Component Analysis (PCA) and k-means clustering, to identify anomalous behaviors in SRF cavities. Our method detects subtle anomalous behavior by analyzing SRF signal data. This knowledge allows for the early detection and resolution of potential faults, significantly improving the efficiency and reliability of operations. Third, we extend these insights to time-series anomaly detection more broadly. We design a contrastive-learning based model tailored to increasingly dynamic environments and academic research. This model improves detection accuracy in settings that require real-time monitoring and predictive maintenance. Our research underscores the broader applicability and impact of advanced machine learning techniques in anomaly detection. By extracting meaningful patterns from complex data, machine learning can significantly enhance security in distributed neural networks and improve the efficiency of particle accelerator operations. This dissertation serves as a stepping stone for future investigations into the vast possibilities of anomaly detection, inspiring further exploration and development of machine learning techniques in this field.

Ferguson, Hal [Old Dominion University]↗

DOE OSTI2026

Enhancing segmentation fairness through curriculum learning and progressive loss: a centralized and federated perspective on radiograph analysis

Bias in medical image segmentation can lead to unequal performance across demographic subgroups, raising concerns about fairness and reliability in clinical AI systems. While deep learning models have achieved high segmentation accuracy, ensuring equitable performance across race and gender remains a significant challenge, particularly in privacy-sensitive healthcare environments. This study investigates fairness-aware medical image segmentation for hip and knee radiographs using deep learning models evaluated in both centralized and Federated Learning (FL) settings. We introduce Curriculum Learning (CL) strategies and Progressive Loss (PL) functions to regulate sample difficulty during training. In addition, we propose two novel fairness-oriented federated learning algorithms, Federated Intersection over Union (FedIoU) and Federated Intersection over Union with Outlier Analysis (FedIoUoutlier). Experiments are conducted using multiple segmentation backbones and simulated multi-site data partitions derived from the Osteoarthritis Initiative dataset. Model performance is evaluated using Intersection over Union (IoU), IoU standard deviation, Skewed Error Ratio (SER), and Min-Max Disparity across race and gender subgroups. Statistical significance was verified using paired t-tests to compare per-sample IoU performance against baseline configurations. Across both hip and knee segmentation tasks, curriculum learning and progressive loss strategies consistently improved segmentation accuracy and reduced demographic performance disparities in centralized training. In federated settings, fairness-aware aggregation further enhanced performance. Notably, FedIoUoutlier combined with balanced curriculum learning and tiered progressive loss achieved the highest mean IoU while yielding the lowest SER and Min-Max Disparity, indicating improved fairness without sacrificing accuracy. In several configurations, federated models matched or exceeded the performance of optimized centralized models, with statistically significant improvements in per-sample IoU over baseline configurations. The results demonstrate that structured training strategies and fairness-aware federated aggregation can jointly improve accuracy, stability, and demographic fairness in medical image segmentation. By integrating curriculum learning, progressive loss, and novel FL algorithms, this work provides a practical pathway toward equitable and privacy-preserving AI systems for medical imaging.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS2015

Evalution of a DE-Identification Process for Ocular Imaging

Medical privacy of NASA astronauts requires an organized and comprehensive approach when data are being made available outside NASA systems. A combination of factors, including the uniquely small patient population, the extensive medical testing done on these individuals, and the relative cultural popularity of the astronauts puts them at a far greater risk to potential exposure of personal information than the general public. Therefore, care must be taken to ensure that the astronauts' identities are concealed. Magnetic Resonance Imaging (MRI) medical data is a recent source of interest to researchers concerned with the development of Visual Impairment due to Intracranial Pressure (VIIP) in the astronaut population. Each vision MRI scan of an astronaut includes 176 separate sagittal images that are saved as an "image series" for clinical use. In addition to the medical information these image sets provide, they also inherently contain a substantial amount of non-medical personally identifiable information (PII) such as-name, date of birth, and date of exam. We have shown that an image set of this type can be rendered, using free software, to give an accurate representation of the patient's face. This currently restricts NASA from dispensing MRI data to researchers in a deidentified format. Automated software programs, such as the Brain Extraction Tool, are available to researchers who wish to de-identify MRI sagittal brain images by "erasing" identifying characteristics such as the nose and jaw on the image sets. However, this software is not useful to NASA for vision research because it removes the portion of the images around the eye orbits, which is the main area of interest to researchers studying the VIIP syndrome. The Lifetime Surveillance of Astronaut Health program has resolved this issue by developing a protocol to de-identify MRI sagittal brain images using Showcase Premier, a DICOM (Digital Imaging and Communications in Medicine) software package. The software allows manual editing of one image from a patient's image set to be automatically applied to the entire image series. This new approach would allow a new level of access to untapped medical imaging data relating to VIIP that can be utilized by researchers while protecting the privacy of the astronauts. In the next step toward finalizing this technique, NASA clinical radiology consultants will test the images to verify removal of all metadata and PII.

LaPelusa, Michael B.↗

NASA NTRS2018

Giving the Public a Perspective into Unmanned Aircraft Systems' Operations

NASA is currently engaged in research to safely enable large-scale commercial applications of small Unmanned Aerial Systems (UAS) in low altitude airspace. This research effort, referred to as UAS Traffic Management (UTM), encompasses the concepts and technologies needed to accommodate the projected demand of UAS operating in the national airspace. One aspect related to the successful implementation of UTM in the future is public acceptance. Transparency will heavily influence this acceptance, and a public portal will provide much of that transparency through ease of access to information about the operations - mainly who, why, and where such operations are taking place. Related concerns to the public are individual privacy, security, and accountability of the operators. Providing the aforementioned information about operations can mitigate these concerns, but a balance will have to be achieved between the need for transparency from the public and the privacy of the operators. The proper balance and the needs of the various UTM stakeholders with regard to information access will be explored through the development and testing of a public portal as part of NASA's Technical Capability 3 (TCL 3) demonstration. Additionally, various approaches to the display of information and user interfaces will be surveyed through the development of a public portal by multiple UTM industry partners across different test sites.

UAS↗

NASA NTRSDate not supplied

Zero-Trust Architecture for Autonomous Edge Computing

We are at the apex of an aviation revolution where autonomy will play a central role in enabling complex, multi-agent systems to communicate, interact, and collaborate on a myriad of applications spanning autonomous swarms to wild-fire management. Autonomy is not an absolute but rather a spectrum ranging from a system requiring significant human intervention to one requiring little to none [1]. For example, the extreme, in the case of an autonomous aircraft, is one that operates independently in the airspace interacting with all other elements (air traffic controllers, other pilots) as if it were a human pilot. Critical to this vision is an architecture that enables autonomous agents to interact with minimal latency. Edge computing is an emerging architecture where compute and storage is pushed to the ‘edge’ of the network in order to minimize the round-trip time from agent to resource thereby mitigating the latency associated with cloud-only based approaches. Additionally, services can generate massive amounts of data (e.g., video feeds), which may require analysis in near real-time. Moving this data to the cloud for further processing may not be feasible due to latency, bandwidth, and cost. Privacy, security, and reliability can also be improved by edge computing architectures. However, this geo-distributed and dynamic* architecture complicates the establishment of unambiguous network security boundaries and can lead to vulnerabilities including man in the middle attacks, replay attacks, physical security breaches of edge nodes, signal interception, etc. This motivates the need for zero-trust architectures [2–4] which de-emphasize the notion of static network perimeters and, as the name implies, do not instill any innate trust in any particular agent. It is required that all agents must be authorized and approved in every transaction. In this paper, we present a zero-trust architecture suitable for edge-computing applications that demand significant low-latency, security, privacy, and reliability.

zero trust↗

DOE OSTI2024

Efficient Anomaly Detection Driven By Different Machine Learning Architectures And Models

The rapid growth and ubiquitous adoption of the internet and cyber-physical systems (CPS) have fundamentally transformed modern communication, work, and human-system interactions. While networks now form the backbone of critical digital ecosystems, enabling seamless data transmission across diverse, interconnected systems, this increased connectivity also expands the attack surface, making real-time detection of network intrusions and anomalies a pressing challenge. Detecting unusual activities within network infrastructure requires advanced data traffic analysis to differentiate between legitimate and malicious interactions. Traditional approaches to network anomaly detectionâ??such as rule-based and signature-based systemsâ??often depend on predefined patterns to identify known anomalies, limiting their effectiveness against emerging, stealthy, or previously unseen threats. These conventional methods suffer from high false alarm rates and fail to adapt to the ever-evolving nature of network traffic, particularly in large-scale, decentralized environments where data volume, velocity, and variety are constantly increasing. This dissertation presents artificial intelligence (AI)-driven approaches to anomaly detection that leverage graphics processing unit (GPU)-enabled high-performance computing (HPC) platforms for processing massive network traffic data and monitoring the components of cyber-physical systems (CPS) for potentially hazardous conditions. The research advances several key contributions: (1) Designing efficient machine learning techniques for CPS condition monitoring and anomaly detection; (2) enabling federated learning (FL) frameworks that enable distributed detection while preserving data privacy and system resilience; (3) exploring graph-based methodologies combining graph neural networks (GNN) and graph machine learning (ML) approaches for the Internet of Things (IoT) and automotive network security, and (4) performing distributed edge computing optimizations that integrate FL with scalable technologies for reduced communication overhead. Through extensive experiments, these methodologies demonstrate that complex anomaly detection and condition monitoring tasks can be achieved while balancing computational efficiency and detection accuracy through fine-grained network information processing. The frameworks developed in this research establish a robust foundation for network anomaly detection, providing scalable, adaptive, and privacy-preserving solutions for safeguarding CPS and IoT networks in an increasingly interconnected digital landscape. The practical implications of these research findings are significant, as they can inform the development of next-generation network security systems and contribute to the protection of critical infrastructure against sophisticated cyber attacks.

Marfo, William↗

DOE OSTI2026

Transportation Secure Data Center: Frequently Asked Questions for Data Owners/Contributors

The Transportation Secure Data Center is a centralized repository for detailed transportation data from travel and transit surveys and studies conducted across the nation. It makes vital transportation data broadly available to users while preserving the privacy of survey participants. Hundreds of datasets from surveys and studies of household travel and transit passenger travel are archived in the TSDC, including surveys and studies conducted by state departments of transportation, metropolitan planning organizations, transit agencies, cities, and other public agencies. Detailed data from travel surveys and studies are extremely valuable for research purposes. However, the fine-grained information they contain could potentially be misused to identify individual travelers, so access to these data should only be granted with safeguards in place to protect participant privacy. The TSDC was created to address this challenge and to relieve public agencies from the burden of archiving their data and responding to data requests.

33 ADVANCED PROPULSION SYSTEMS↗

DOE OSTI2025

Enabling end-to-end secure federated learning in biomedical research on heterogeneous computing environments with APPFLx

Facilitating large-scale, cross-institutional collaboration in biomedical machine learning (ML) projects requires a trustworthy and resilient federated learning (FL) environment to ensure that sensitive information such as protected health information is kept confidential. Specifically designed for this purpose, this work introduces APPFLx - a low-code, easy-to-use FL framework that enables easy setup, configuration, and running of FL experiments. APPFLx removes administrative boundaries of research organizations and healthcare systems while providing secure end-to-end communication, privacy-preserving functionality, and identity management. Furthermore, it is completely agnostic to the underlying computational infrastructure of participating clients, allowing an instantaneous deployment of this framework into existing computing infrastructures. Experimentally, the utility of APPFLx is demonstrated in two case studies: (1) predicting participant age from electrocardiogram (ECG) waveforms, and (2) detecting COVID-19 disease from chest radiographs. Here, ML models were securely trained across heterogeneous computing resources, including a combination of on-premise high-performance computing and cloud computing facilities. By securely unlocking data from multiple sources for training without directly sharing it, these FL models enhance generalizability and performance compared to centralized training models while ensuring data remains protected. In conclusion, APPFLx demonstrated itself as an easy-to-use framework for accelerating biomedical studies across organizations and healthcare systems on large datasets while maintaining the protection of private medical data.

Biomedical Research↗

DOE OSTI2025

Towards the next generation of Geospatial Artificial Intelligence

Geospatial Artificial Intelligence (GeoAI), as the integration of geospatial studies and AI, has become one of the fastest-developing research directions in spatial data science and geography. This rapid change in the field calls for a deeper understanding of the recent developments and envision where the field is going in the near future. In this work, we provide a quantitative analysis of the GeoAI literature from the spatial, temporal, and semantic aspects. We briefly discuss the history of AI and GeoAI by highlighting some pioneering work. Then we discuss the current landscape of GeoAI by selecting five representative subdomains including remote sensing, urban computing, Earth system science, cartography, and geospatial semantics. Finally, we highlight several unique future research directions of GeoAI which are classified into two groups: GeoAI method development challenges and GeoAI Ethics challenges. Topics include heterogeneity-aware GeoAI, knowledge-guided GeoAI, spatial representation learning, geo-foundation models, fairness-aware GeoAI, privacy-aware GeoAI, as well as interpretable and explainable GeoAI. We hope our review of GeoAI’s past, present, and future is comprehensive and can enlighten the next generation of GeoAI research.

58 GEOSCIENCES↗

DOE OSTI2024

An Empirical Validation of a Constrained Bin Packing Algorithm for a Home Energy Management System

The increasing number of intelligent electrical appliances and home energy management systems provide a big opportunity for demand response services from residential and small commercial buildings to the grid. Simultaneously, direct control of individual devices by utilities can cause communication bottlenecks, as well as coordination and privacy concerns. These challenges can be addressed by combining the constituent devices into a single house battery equivalent for the purposes of demand response, using Minkowski sum and a 2d bin packing problem. However, the well-studied traditional problems have not been tested in a real house, as implementation carries significant challenges of its own. We deploy the packing problem on residential devices in a controllable house. We report the barriers we found, such as charge forecast and scalability of the algorithm, and discuss our solutions. The study serves as an intermediate step between existing theoretical research and possible future steps, such as prototype deployments of systems that provide residential demand response.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2026

Generative AI in Supply Chain Management: Applications, Challenges, and Future Directions

Supply chain management (SCM) is undergoing rapid transformation due to increasing global complexity, demand volatility, and operational disruptions. Generative Artificial Intelligence (GenAI) has emerged as a powerful paradigm capable of synthesizing data, simulating operational scenarios, and enabling adaptive decision-making across supply chain networks. This paper presents a survey of GenAI’s role in SCM, focusing on its applications in predictive analytics, autonomous logistics, and fraud detection. Unlike traditional AI systems that rely primarily on predictive analytics, GenAI models, including large language models, generative adversarial networks, and diffusion-based architectures, enable the creation of synthetic supply chain scenarios and autonomous optimization strategies. This survey provides (1) a taxonomy of GenAI techniques for supply chain applications, (2) a comparative analysis of generative AI approaches with traditional machine learning, reinforcement learning, and blockchain-based methods, and (3) a discussion of key challenges such as data privacy, interpretability, and integration with legacy enterprise systems. Furthermore, we outline open research problems and propose directions for future research toward autonomous, resilient, and sustainable AI-driven supply chains.

15 - GEOTHERMAL ENERGY↗

DOE OSTI2026

FIRM: federated image reconstruction using multimodal tomographic data

Here, we propose a federated algorithm for reconstructing images using multimodal tomographic data sourced from dispersed locations, addressing the challenges of traditional unimodal approaches that are prone to noise and reduced image quality, as well as the limitations of centralized multimodal approaches that require extensive data transfer, leading to significant communication overhead, storage demands, and potential data privacy concerns. Our approach formulates a joint inverse optimization problem incorporating multimodality constraints and solves it in a federated framework through local gradient computations complemented by lightweight central operations, thereby ensuring data decentralization. Leveraging the connection between our federated algorithm and the quadratic penalty method, we introduce an adaptive step-size rule with guaranteed sublinear convergence. Numerical results demonstrate superior computational efficiency and improved image reconstruction quality compared to existing approaches.

federated algorithm↗

DOE OSTI2024

chatHPC: Empowering HPC users with large language models

The ever-growing number of pre-trained large language models (LLMs) across scientific domains presents a challenge for application developers. While these models offer vast potential, fine-tuning them with custom data, aligning them for specific tasks, and evaluating their performance remain crucial steps for effective utilization. However, applying these techniques to models with tens of billions of parameters can take days or even weeks on modern workstations, making the cumulative cost of model comparison and evaluation a significant barrier to LLM-based application development. To address this challenge, we introduce an end-to-end pipeline specifically designed for building conversational and programmable AI agents on high performance computing (HPC) platforms. Our comprehensive pipeline encompasses: model pre-training, fine-tuning, web and API service deployment, along with crucial evaluations for lexical coherence, semantic accuracy, hallucination detection, and privacy considerations. Here, we demonstrate our pipeline through the development of chatHPC, a chatbot for HPC question answering and script generation. Leveraging our scalable pipeline, we achieve end-to-end LLM alignment in under an hour on the Frontier supercomputer. We propose a novel self-improved, self-instruction method for instruction set generation, investigate scaling and fine-tuning strategies, and conduct a systematic evaluation of model performance. The established practices within chatHPC will serve as a valuable guidance for future LLM-based application development on HPC platforms.

97 MATHEMATICS AND COMPUTING↗