arXiv subjects

Shigang Li

Publications and source records attributed to Shigang Li.

At least 19 recordsLinked to original sources

arXiv2026

ParamSpMM: Adaptive and Efficient Sparse Matrix-Matrix Multiplication on GPUs for GNNs

Fueled by the ability to mine real-world graph data, GNN applications have experienced phenomenal growth. Sparse Matrix-Matrix Multiplication (SpMM) is a critical operator in GNNs. However, existing SpMM designs for GNNs struggle to adapt to diverse input characteristics. In this paper, we first conduct a comprehensive analysis of existing SpMM optimizations, revealing their limitations through statistical and empirical evidence. Based on this analysis, we introduce ParamSpMM, a parametric approach for highly adaptive and efficient SpMM computation in GNNs. It incorporates a new data structure, the Parameterized Compressed Sparse Row (PCSR), to flexibly integrate existing optimization techniques. ParamSpMM enables the configuration of these optimization techniques according to various input characteristics. Furthermore, we complement ParamSpMM with an ML-based SpMM-decider that predicts optimal configurations based on carefully crafted input features. Our evaluations demonstrate that ParamSpMM outperforms Nvidia cuSPARSE with an average speedup of 1.92x, significantly enhancing GNN training efficiency.

cs.DC↗

arXiv2025

Dual-Projection Fusion for Accurate Upright Panorama Generation in Robotic Vision

Panoramic cameras, capable of capturing a 360-degree field of view, are crucial in robotic vision, particularly in environments with sparse features. However, non-upright panoramas due to unstable robot postures hinder downstream tasks. Traditional IMU-based correction methods suffer from drift and external disturbances, while vision-based approaches offer a promising alternative. This study presents a dual-stream angle-aware generation network that jointly estimates camera inclination angles and reconstructs upright panoramic images. The network comprises a CNN branch that extracts local geometric structures from equirectangular projections and a ViT branch that captures global contextual cues from cubemap projections. These are integrated through a dual-projection adaptive fusion module that aligns spatial features across both domains. To further enhance performance, we introduce a high-frequency enhancement block, circular padding, and channel attention mechanisms to preserve 360{\deg} continuity and improve geometric sensitivity. Experiments on the SUN360 and M3D datasets demonstrate that our method outperforms existing approaches in both inclination estimation and upright panorama generation. Ablation studies further validate the contribution of each module and highlight the synergy between the two tasks. The code and related datasets can be found at: https://github.com/YuhaoShine/DualProjectionFusion.

cs.CV↗

arXiv2025

BitSnap: Checkpoint Sparsification and Quantization in LLM Training

As large language models (LLMs) continue to grow in size and complexity, efficient checkpoint saving\&loading has become crucial for managing storage, memory usage, and fault tolerance in LLM training. The current works do not comprehensively take into account the optimization of these several aspects. This paper proposes a novel checkpoint sparsification and quantization method that adapts dynamically to different training stages and model architectures. We present a comprehensive analysis of existing lossy and lossless compression techniques, identify current limitations, and introduce our adaptive approach that balances compression ratio, speed, and precision impact throughout the training process. Experiments on different sizes of LLMs demonstrate that our bitmask-based sparsification method achieves 16x compression ratio without compromising model accuracy. Additionally, the cluster-based quantization method achieves 2x compression ratio with little precision loss.

cs.LG↗

arXiv2025

360-Degree Full-view Image Segmentation by Spherical Convolution compatible with Large-scale Planar Pre-trained Models

Due to the current lack of large-scale datasets at the million-scale level, tasks involving panoramic images predominantly rely on existing two-dimensional pre-trained image benchmark models as backbone networks. However, these networks are not equipped to recognize the distortions and discontinuities inherent in panoramic images, which adversely affects their performance in such tasks. In this paper, we introduce a novel spherical sampling method for panoramic images that enables the direct utilization of existing pre-trained models developed for two-dimensional images. Our method employs spherical discrete sampling based on the weights of the pre-trained models, effectively mitigating distortions while achieving favorable initial training values. Additionally, we apply the proposed sampling method to panoramic image segmentation, utilizing features obtained from the spherical model as masks for specific channel attentions, which yields commendable results on commonly used indoor datasets, Stanford2D3D.

cs.CV↗

arXiv2025

Libra: Unleashing GPU Heterogeneity for High-Performance Sparse Matrix Multiplication

Sparse matrix multiplication operators (i.e., SpMM and SDDMM) are widely used in deep learning and scientific computing. Modern accelerators are commonly equipped with Tensor Core Units (TCUs) and CUDA cores to accelerate sparse operators. The former excels at structured matrix computations, whereas the latter offers greater programming flexibility. However, how to combine these two resources to maximize sparse-operator performance remains unclear. In this work, we first identify the source of performance gains in hybrid computation and systematically analyze their complementary strengths. Motivated by this, we propose Libra, a holistic framework that efficiently leverages heterogeneous computing resources to accelerate both SpMM and SDDMM operators. Specifically, Libra introduces a 2D-aware (locality and utilization) workload distribution method to precisely identify the optimal task mapping, simultaneously leveraging the data reuse capabilities of TCUs and the flexibility of CUDA cores to minimize computational redundancy. Libra further incorporates hybrid load balancing, occupancy-aware task scheduling, and efficient kernel implementations to maximize execution efficiency. Extensive experiments on H100 and RTX 4090 GPUs demonstrate that Libra surpasses all the 12 up-to-date baselines significantly, e.g., on average 1.77x speedup over FlashSparse, 1.73x over RoDe, and 2.9x over DGL for end-to-end GNN applications. Libra opens up a new perspective for sparse operator acceleration by fully unleashing the power of heterogeneous GPU resources.

cs.DC↗

arXiv2025

SparkAttention: High-Performance Multi-Head Attention for Large Models on Volta GPU Architecture

Transformer are widely used in various fields such as natural language processing and computer vision. However, the training time for large Transformer models can be challenging due to the Multi-Head Attention (MHA) mechanism. Especially as models become larger, training becomes more costly. So it is crucial to utilize various resources for efficient model training. Currently, NVIDIA Volta GPU is still widely used. However, because the computational shapes supported by Tensor Core Units (TCU) of Volta GPU differ from other GPU architectures, most efforts have not focused on using them to accelerate Transformer training. To address this issue, we propose SparkAttention, an acceleration library designed to speed up MHA training on the Volta GPU. SparkAttention leverages TCU and kernel fusion to reduce the number of high bandwidth memory (HBM) accesses and overhead. Our End-to-End experimental results on an NVIDIA V100 GPU show that SparkAttention achieves on average 1.80$\times$ (up to 2.46$\times$) speedup compared to using PyTorch.

cs.DC↗

arXiv2024

FlashSparse: Minimizing Computation Redundancy for Fast Sparse Matrix Multiplications on Tensor Cores

Sparse Matrix-matrix Multiplication (SpMM) and Sampled Dense-dense Matrix Multiplication (SDDMM) are important sparse operators in scientific computing and deep learning. Tensor Core Units (TCUs) enhance modern accelerators with superior computing power, which is promising to boost the performance of matrix operators to a higher level. However, due to the irregularity of unstructured sparse data, it is difficult to deliver practical speedups on TCUs. To this end, we propose FlashSparse, a novel approach to bridge the gap between sparse workloads and the TCU architecture. Specifically, FlashSparse minimizes the sparse granularity for SpMM and SDDMM on TCUs through a novel swap-and-transpose matrix multiplication strategy. Benefiting from the minimum sparse granularity, the computation redundancy is remarkably reduced while the computing power of TCUs is fully utilized. Besides, FlashSparse is equipped with a memory-efficient thread mapping strategy for coalesced data access and a sparse matrix storage format to save memory footprint. Extensive experimental results on H100 and RTX 4090 GPUs show that FlashSparse sets a new state-of-the-art for sparse matrix multiplications (geometric mean 5.5x speedup over DTC-SpMM and 3.22x speedup over RoDe).

cs.DC↗

arXiv2024

Estimating Depth of Monocular Panoramic Image with Teacher-Student Model Fusing Equirectangular and Spherical Representations

Disconnectivity and distortion are the two problems which must be coped with when processing 360 degrees equirectangular images. In this paper, we propose a method of estimating the depth of monocular panoramic image with a teacher-student model fusing equirectangular and spherical representations. In contrast with the existing methods fusing an equirectangular representation with a cube map representation or tangent representation, a spherical representation is a better choice because a sampling on a sphere is more uniform and can also cope with distortion more effectively. In this processing, a novel spherical convolution kernel computing with sampling points on a sphere is developed to extract features from the spherical representation, and then, a Segmentation Feature Fusion(SFF) methodology is utilized to combine the features with ones extracted from the equirectangular representation. In contrast with the existing methods using a teacher-student model to obtain a lighter model of depth estimation, we use a teacher-student model to learn the latent features of depth images. This results in a trained model which estimates the depth map of an equirectangular image using not only the feature maps extracted from an input equirectangular image but also the distilled knowledge learnt from the ground truth of depth map of a training set. In experiments, the proposed method is tested on several well-known 360 monocular depth estimation benchmark datasets, and outperforms the existing methods for the most evaluation indexes.

cs.CV↗

arXiv2023

TRANSOM: An Efficient Fault-Tolerant System for Training LLMs

Large language models (LLMs) with hundreds of billions or trillions of parameters, represented by chatGPT, have achieved profound impact on various fields. However, training LLMs with super-large-scale parameters requires large high-performance GPU clusters and long training periods lasting for months. Due to the inevitable hardware and software failures in large-scale clusters, maintaining uninterrupted and long-duration training is extremely challenging. As a result, A substantial amount of training time is devoted to task checkpoint saving and loading, task rescheduling and restart, and task manual anomaly checks, which greatly harms the overall training efficiency. To address these issues, we propose TRANSOM, a novel fault-tolerant LLM training system. In this work, we design three key subsystems: the training pipeline automatic fault tolerance and recovery mechanism named Transom Operator and Launcher (TOL), the training task multi-dimensional metric automatic anomaly detection system named Transom Eagle Eye (TEE), and the training checkpoint asynchronous access automatic fault tolerance and recovery technology named Transom Checkpoint Engine (TCE). Here, TOL manages the lifecycle of training tasks, while TEE is responsible for task monitoring and anomaly reporting. TEE detects training anomalies and reports them to TOL, who automatically enters the fault tolerance strategy to eliminate abnormal nodes and restart the training task. And the asynchronous checkpoint saving and loading functionality provided by TCE greatly shorten the fault tolerance overhead. The experimental results indicate that TRANSOM significantly enhances the efficiency of large-scale LLM training on clusters. Specifically, the pre-training time for GPT3-175B has been reduced by 28%, while checkpoint saving and loading performance have improved by a factor of 20.

cs.DC↗

arXiv2023

A High-Performance Design, Implementation, Deployment, and Evaluation of The Slim Fly Network

Novel low-diameter network topologies such as Slim Fly (SF) offer significant cost and power advantages over the established Fat Tree, Clos, or Dragonfly. To spearhead the adoption of low-diameter networks, we design, implement, deploy, and evaluate the first real-world SF installation. We focus on deployment, management, and operational aspects of our test cluster with 200 servers and carefully analyze performance. We demonstrate techniques for simple cabling and cabling validation as well as a novel high-performance routing architecture for InfiniBand-based low-diameter topologies. Our real-world benchmarks show SF's strong performance for many modern workloads such as deep neural network training, graph analytics, or linear algebra kernels. SF outperforms non-blocking Fat Trees in scalability while offering comparable or better performance and lower cost for large network sizes. Our work can facilitate deploying SF while the associated (open-source) routing architecture is fully portable and applicable to accelerate any low-diameter interconnect.

cs.NI↗

arXiv2023

Co-design Hardware and Algorithm for Vector Search

Vector search has emerged as the foundation for large-scale information retrieval and machine learning systems, with search engines like Google and Bing processing tens of thousands of queries per second on petabyte-scale document datasets by evaluating vector similarities between encoded query texts and web documents. As performance demands for vector search systems surge, accelerated hardware offers a promising solution in the post-Moore's Law era. We introduce \textit{FANNS}, an end-to-end and scalable vector search framework on FPGAs. Given a user-provided recall requirement on a dataset and a hardware resource budget, \textit{FANNS} automatically co-designs hardware and algorithm, subsequently generating the corresponding accelerator. The framework also supports scale-out by incorporating a hardware TCP/IP stack in the accelerator. \textit{FANNS} attains up to 23.0$\times$ and 37.2$\times$ speedup compared to FPGA and CPU baselines, respectively, and demonstrates superior scalability to GPUs, achieving 5.5$\times$ and 7.6$\times$ speedup in median and 95\textsuperscript{th} percentile (P95) latency within an eight-accelerator configuration. The remarkable performance of \textit{FANNS} lays a robust groundwork for future FPGA integration in data centers and AI supercomputers.

cs.LG↗

arXiv2023

ASDL: A Unified Interface for Gradient Preconditioning in PyTorch

Gradient preconditioning is a key technique to integrate the second-order information into gradients for improving and extending gradient-based learning algorithms. In deep learning, stochasticity, nonconvexity, and high dimensionality lead to a wide variety of gradient preconditioning methods, with implementation complexity and inconsistent performance and feasibility. We propose the Automatic Second-order Differentiation Library (ASDL), an extension library for PyTorch, which offers various implementations and a plug-and-play unified interface for gradient preconditioning. ASDL enables the study and structured comparison of a range of gradient preconditioning methods.

cs.LG↗

arXiv2023

An End-to-End Network for Upright Adjustment of Panoramic Images

Nowadays, panoramic images can be easily obtained by panoramic cameras. However, when the panoramic camera orientation is tilted, a non-upright panoramic image will be captured. Existing upright adjustment models focus on how to estimate more accurate camera orientation, and attribute image reconstruction to offline or post-processing tasks. To this end, we propose an online end-to-end network for upright adjustment. Our network is designed to reconstruct the image while finding the angle. Our network consists of three modules: orientation estimation, LUT online generation, and upright reconstruction. Direction estimation estimates the tilt angle of the panoramic image. Then, a converter block with upsampling function is designed to generate angle to LUT. This module can output corresponding online LUT for different input angles. Finally, a lightweight generative adversarial network (GAN) aims to generate upright images from shallow features. The experimental results show that in terms of angles, we have improved the accuracy of small angle errors. In terms of image reconstruction, In image reconstruction, we have achieved the first real-time online upright reconstruction of panoramic images using deep learning networks.

cs.CV↗

arXiv2023

AutoDDL: Automatic Distributed Deep Learning with Near-Optimal Bandwidth Cost

Recent advances in deep learning are driven by the growing scale of computation, data, and models. However, efficiently training large-scale models on distributed systems requires an intricate combination of data, operator, and pipeline parallelism, which exerts heavy burden on machine learning practitioners. To this end, we propose AutoDDL, a distributed training framework that automatically explores and exploits new parallelization schemes with near-optimal bandwidth cost. AutoDDL facilitates the description and implementation of different schemes by utilizing OneFlow's Split, Broadcast, and Partial Sum (SBP) abstraction. AutoDDL is equipped with an analytical performance model combined with a customized Coordinate Descent algorithm, which significantly reduces the scheme searching overhead. We conduct evaluations on Multi-Node-Single-GPU and Multi-Node-Multi-GPU machines using different models, including VGG and Transformer. Compared to the expert-optimized implementations, AutoDDL reduces the end-to-end training time by up to 31.1% and 10% for Transformer and up to 17.7% and 71.5% for VGG on the two parallel systems, respectively.

cs.DC↗

arXiv2022

PipeFisher: Efficient Training of Large Language Models Using Pipelining and Fisher Information Matrices

Pipeline parallelism enables efficient training of Large Language Models (LLMs) on large-scale distributed accelerator clusters. Yet, pipeline bubbles during startup and tear-down reduce the utilization of accelerators. Although efficient pipeline schemes with micro-batching and bidirectional pipelines have been proposed to maximize utilization, a significant number of bubbles cannot be filled using synchronous forward and backward passes. To address this problem, we suggest that extra work be assigned to the bubbles to gain auxiliary benefits in LLM training. As an example in this direction, we propose PipeFisher, which assigns the work of K-FAC, a second-order optimization method based on the Fisher information matrix, to the bubbles to accelerate convergence. In Phase 1 pretraining of BERT-Base and -Large models, PipeFisher reduces the (simulated) training time to 50-75% compared to training with a first-order optimizer by greatly improving the accelerator utilization and benefiting from the improved convergence by K-FAC.

cs.LG↗

arXiv2022

HammingMesh: A Network Topology for Large-Scale Deep Learning

Numerous microarchitectural optimizations unlocked tremendous processing power for deep neural networks that in turn fueled the AI revolution. With the exhaustion of such optimizations, the growth of modern AI is now gated by the performance of training systems, especially their data movement. Instead of focusing on single accelerators, we investigate data-movement characteristics of large-scale training at full system scale. Based on our workload analysis, we design HammingMesh, a novel network topology that provides high bandwidth at low cost with high job scheduling flexibility. Specifically, HammingMesh can support full bandwidth and isolation to deep learning training jobs with two dimensions of parallelism. Furthermore, it also supports high global bandwidth for generic traffic. Thus, HammingMesh will power future large-scale deep learning systems with extreme bandwidth requirements.

cs.DC↗

arXiv2022

Efficient Quantized Sparse Matrix Operations on Tensor Cores

The exponentially growing model size drives the continued success of deep learning, but it brings prohibitive computation and memory cost. From the algorithm perspective, model sparsification and quantization have been studied to alleviate the problem. From the architecture perspective, hardware vendors provide Tensor cores for acceleration. However, it is very challenging to gain practical speedups from sparse, low-precision matrix operations on Tensor cores, because of the strict requirements for data layout and lack of support for efficiently manipulating the low-precision integers. We propose Magicube, a high-performance sparse-matrix library for low-precision integers on Tensor cores. Magicube supports SpMM and SDDMM, two major sparse operations in deep learning with mixed precision. Experimental results on an NVIDIA A100 GPU show that Magicube achieves on average 1.44x (up to 2.37x) speedup over the vendor-optimized library for sparse kernels, and 1.43x speedup over the state-of-the-art with a comparable accuracy for end-to-end sparse Transformer inference.

cs.DC↗

arXiv2022

A Data-Centric Optimization Framework for Machine Learning

Rapid progress in deep learning is leading to a diverse set of quickly changing models, with a dramatically growing demand for compute. However, as frameworks specialize performance optimization to patterns in popular networks, they implicitly constrain novel and diverse models that drive progress in research. We empower deep learning researchers by defining a flexible and user-customizable pipeline for optimizing training of arbitrary deep neural networks, based on data movement minimization. The pipeline begins with standard networks in PyTorch or ONNX and transforms computation through progressive lowering. We define four levels of general-purpose transformations, from local intra-operator optimizations to global data movement reduction. These operate on a data-centric graph intermediate representation that expresses computation and data movement at all levels of abstraction, including expanding basic operators such as convolutions to their underlying computations. Central to the design is the interactive and introspectable nature of the pipeline. Every part is extensible through a Python API, and can be tuned interactively using a GUI. We demonstrate competitive performance or speedups on ten different networks, with interactive optimizations discovering new opportunities in EfficientNet.

cs.LG↗