arXiv subjects

Shenggan Cheng

Publications and source records attributed to Shenggan Cheng.

17 recordsLinked to original sources

arXiv2026

Transforming the Use of Earth Observation Data: Exascale Training of a Generative Compression Model with Historical Priors for up to 10,000x Data Reduction

Earth observation is becoming one of the largest data-producing activities in science, yet current pipelines still treat compression as a storage and transmission tool rather than a new way to use data. We present a generative compression framework that learns from historical Earth observation archives and enables on-demand 100x to 10,000x data reduction across downstream tasks. Unlike general visual data, Earth observation repeatedly measures the same evolving planet, making historical-prior learning feasible for extreme compression. To realize this paradigm, we train large generative compression models at exascale on the LineShine Armv9 CPU supercomputer, with co-optimization across model design, kernels, memory hierarchy, runtime, and parallelism. Our implementation sustains 1.54 EFLOP/s and peaks at 2.16 EFLOP/s in end-to-end training. This work shows that historical-prior generative compression can turn Earth observation data into an active, task-adaptive foundation for acquisition, delivery, storage, and scientific use.

cs.DC↗

arXiv2026

DiT-HC: Enabling Efficient Training of Visual Generation Model DiT on HPC-oriented CPU Cluster

Generative foundation models have become an important tool for data reconstruction and simulation in scientific computing, showing a tight integration with traditional numerical simulations. At the same time, with the development of new hardware features, such as matrix acceleration units and high-bandwidth memory, CPU-based clusters offer promising opportunities to accelerate and scale such models, facilitating the unification of artificial intelligence and scientific computing. We present DiT-HC, the first system to train and scale the generative model DiT on a next-generation HPC CPU cluster. DiT-HC introduces three key techniques: (1) communication-free tensor parallelism (CFTP) with AutoMem for automated memory-aware dataflow, (2) HCOps, a suite of optimized GEMM and operator kernels leveraging vector and matrix acceleration units, and (3) a custom MPI backend that overlaps computation, communication, and memory movement. Experiments show 8.2 to 87.7 times speedups over native or public CPU libraries and 90.6% weak scaling efficiency on 256 nodes. These results demonstrate the feasibility of large-scale generative model training on CPU clusters and provide new insights for future HPC-AI co-design.

cs.DC↗

arXiv2025

Expert-as-a-Service: Towards Efficient, Scalable, and Robust Large-scale MoE Serving

Mixture-of-Experts (MoE) models challenge serving infrastructures with dynamic, sparse expert utilization, causing instability on conventional systems designed for dense architectures. We propose EaaS, a novel serving system to enable efficient, scalable, and robust MoE deployment. Our system disaggregates MoE modules into independent, stateless services. This design enables fine-grained resource scaling and provides inherent fault tolerance by decoupling compute units. The architecture is powered by a high-performance, CPU-free peer-to-peer communication library that ensures minimal overhead and high throughput. Experiments confirm EaaS's scalability and efficiency, achieving performance comparable to monolithic systems while providing robust fault tolerance and strong scalability. EaaS incurs less than a 2% throughput reduction under simulated hardware failures that would otherwise halt monolithic architectures. It further saves up to 37.5% of computing resources through dynamic fine-grained adaptation to serving traffic, demonstrating strong resilience for large-scale MoE deployment in production.

cs.DC↗

arXiv2025

Adaptive Hybrid Caching for Efficient Text-to-Video Diffusion Model Acceleration

Efficient video generation models are increasingly vital for multimedia synthetic content generation. Leveraging the Transformer architecture and the diffusion process, video DiT models have emerged as a dominant approach for high-quality video generation. However, their multi-step iterative denoising process incurs high computational cost and inference latency. Caching, a widely adopted optimization method in DiT models, leverages the redundancy in the diffusion process to skip computations in different granularities (e.g., step, cfg, block). Nevertheless, existing caching methods are limited to single-granularity strategies, struggling to balance generation quality and inference speed in a flexible manner. In this work, we propose MixCache, a training-free caching-based framework for efficient video DiT inference. It first distinguishes the interference and boundary between different caching strategies, and then introduces a context-aware cache triggering strategy to determine when caching should be enabled, along with an adaptive hybrid cache decision strategy for dynamically selecting the optimal caching granularity. Extensive experiments on diverse models demonstrate that, MixCache can significantly accelerate video generation (e.g., 1.94$\times$ speedup on Wan 14B, 1.97$\times$ speedup on HunyuanVideo) while delivering both superior generation quality and inference efficiency compared to baseline methods.

cs.GR↗

arXiv2025

ElasticMM: Efficient Multimodal LLMs Serving with Elastic Multimodal Parallelism

Multimodal large language models (MLLMs) extend LLMs to handle images, videos, and audio by incorporating feature extractors and projection modules. However, these additional components -- combined with complex inference pipelines and heterogeneous workloads -- introduce significant inference overhead. Therefore, efficiently serving MLLMs remains a major challenge. Current tightly coupled serving architectures struggle to distinguish between mixed request types or adapt parallelism strategies to different inference stages, leading to increased time-to-first-token (TTFT) latency and poor resource utilization. To address this, we introduce Elastic Multimodal Parallelism (EMP), a new serving paradigm that elastically adapts to resource heterogeneity across request types and inference stages. Building upon EMP, we develop ElasticMM, an MLLM serving system that (1) separates requests into independent modality groups with dynamic resource allocation via a modality-aware load balancer; (2) decouples inference stages and enables parallelism adjustment and adaptive scaling via elastic partition scheduling; and (3) improves inference efficiency through unified multimodal prefix caching and non-blocking encoding. Experiments on diverse real-world datasets show that ElasticMM outperforms state-of-the-art (SOTA) serving systems, reducing TTFT by up to 4.2x and achieving 3.2-4.5x higher throughput while meeting service-level objectives (SLOs).

cs.DC↗

arXiv2025

HelixPipe: Efficient Distributed Training of Long Sequence Transformers with Attention Parallel Pipeline Parallelism

As transformer sequence lengths grow, existing pipeline parallelisms incur suboptimal performance due to the quadratic attention computation and the substantial memory overhead. To relieve these challenges, we propose HelixPipe, a novel pipeline parallelism for long sequence transformer training. First, HelixPipe introduces attention parallel partition, which schedules attention computations of different micro batches across different pipeline stages in parallel, reducing pipeline bubbles. Second, it employs a two-fold first-in-last-out micro batch schedule to balance memory usage and overlap communication with computation. Additionally, HelixPipe utilizes recomputation without attention and chunked MLP to mitigate fragmentation and enable longer sequences. Experiments demonstrate that HelixPipe gains increasing advantages with longer sequence lengths, and outperforms existing methods in throughput and scalability across varying pipeline sizes, model sizes, and cluster configurations. Notably, it achieves a 26\% speedup over baseline methods when training a 7B model with 128k sequence length on 64 H20 GPUs. Code is available at https://github.com/code-tunnel/Megatron-LM/tree/dev.

cs.LG↗

arXiv2025

SRDiffusion: Accelerate Video Diffusion Inference via Sketching-Rendering Cooperation

Leveraging the diffusion transformer (DiT) architecture, models like Sora, CogVideoX and Wan have achieved remarkable progress in text-to-video, image-to-video, and video editing tasks. Despite these advances, diffusion-based video generation remains computationally intensive, especially for high-resolution, long-duration videos. Prior work accelerates its inference by skipping computation, usually at the cost of severe quality degradation. In this paper, we propose SRDiffusion, a novel framework that leverages collaboration between large and small models to reduce inference cost. The large model handles high-noise steps to ensure semantic and motion fidelity (Sketching), while the smaller model refines visual details in low-noise steps (Rendering). Experimental results demonstrate that our method outperforms existing approaches, over 3$\times$ speedup for Wan with nearly no quality loss for VBench, and 2$\times$ speedup for CogVideoX. Our method is introduced as a new direction orthogonal to existing acceleration strategies, offering a practical solution for scalable video generation.

cs.GR↗

arXiv2024

StarTrail: Concentric Ring Sequence Parallelism for Efficient Near-Infinite-Context Transformer Model Training

Training Transformer models on long sequences in a distributed setting poses significant challenges in terms of efficiency and scalability. Current methods are either constrained by the number of attention heads or excessive communication overheads. To address this problem, we propose StarTrail, a multi-dimensional concentric distributed training system for long sequences, fostering an efficient communication paradigm and providing additional tuning flexibility for communication arrangements. Specifically, StarTrail introduces an extra parallel dimension and divides the peer-to-peer communication into sub-rings to substantially reduce communication volume and avoid bandwidth bottlenecks. Through comprehensive experiments across diverse hardware environments and on both Natural Language Processing (NLP) and Computer Vision (CV) tasks, we demonstrate that our approach significantly surpasses state-of-the-art methods that support Long sequence lengths, achieving performance improvements of up to 77.12% on GPT-style models and up to 114.33% on DiT (Diffusion Transformer) models without affecting the computations results.

cs.DC↗

arXiv2024

DSP: Dynamic Sequence Parallelism for Multi-Dimensional Transformers

Scaling multi-dimensional transformers to long sequences is indispensable across various domains. However, the challenges of large memory requirements and slow speeds of such sequences necessitate sequence parallelism. All existing approaches fall under the category of embedded sequence parallelism, which are limited to shard along a single sequence dimension, thereby introducing significant communication overhead. However, the nature of multi-dimensional transformers involves independent calculations across multiple sequence dimensions. To this end, we propose Dynamic Sequence Parallelism (DSP) as a novel abstraction of sequence parallelism. DSP dynamically switches the parallel dimension among all sequences according to the computation stage with efficient resharding strategy. DSP offers significant reductions in communication costs, adaptability across modules, and ease of implementation with minimal constraints. Experimental evaluations demonstrate DSP's superiority over state-of-the-art embedded sequence parallelism methods by remarkable throughput improvements ranging from 32.2% to 10x, with less than 25% communication volume.

cs.DC↗

arXiv2024

HeteGen: Heterogeneous Parallel Inference for Large Language Models on Resource-Constrained Devices

In recent times, the emergence of Large Language Models (LLMs) has resulted in increasingly larger model size, posing challenges for inference on low-resource devices. Prior approaches have explored offloading to facilitate low-memory inference but often suffer from efficiency due to I/O bottlenecks. To achieve low-latency LLMs inference on resource-constrained devices, we introduce HeteGen, a novel approach that presents a principled framework for heterogeneous parallel computing using CPUs and GPUs. Based on this framework, HeteGen further employs heterogeneous parallel computing and asynchronous overlap for LLMs to mitigate I/O bottlenecks. Our experiments demonstrate a substantial improvement in inference speed, surpassing state-of-the-art methods by over 317% at most.

cs.PF↗

arXiv2024

AutoChunk: Automated Activation Chunk for Memory-Efficient Long Sequence Inference

Large deep learning models have achieved impressive performance across a range of applications. However, their large memory requirements, including parameter memory and activation memory, have become a significant challenge for their practical serving. While existing methods mainly address parameter memory, the importance of activation memory has been overlooked. Especially for long input sequences, activation memory is expected to experience a significant exponential growth as the length of sequences increases. In this approach, we propose AutoChunk, an automatic and adaptive compiler system that efficiently reduces activation memory for long sequence inference by chunk strategies. The proposed system generates chunk plans by optimizing through multiple stages. In each stage, the chunk search pass explores all possible chunk candidates and the chunk selection pass identifies the optimal one. At runtime, AutoChunk employs code generation to automatically apply chunk strategies. The experiments demonstrate that AutoChunk can reduce over 80\% of activation memory while maintaining speed loss within 10%, extend max sequence length by 3.2x to 11.7x, and outperform state-of-the-art methods by a large margin.

cs.PF↗

arXiv2023

Hanayo: Harnessing Wave-like Pipeline Parallelism for Enhanced Large Model Training Efficiency

Large-scale language models have become increasingly challenging and expensive to train. Among various methods addressing this issue, Pipeline Parallelism has been widely employed to accommodate massive model weights within limited GPU memory. This paper introduces Hanayo, a wave-like pipeline parallelism strategy that boasts a concise structure and practical applicability, alongside a high-performance pipeline execution runtime to tackle the challenges of pipeline strategy implementation. Hanayo mitigates the issues of pipeline bubbles and excessive memory consumption prevalent in existing schemes, without resorting to model duplicates as in Chimera. Our evaluation, conducted on four distinct computing clusters and involving both GPT-like and BERT-like architectures with up to 32 GPUs, demonstrates up to a 30.4 \% increase in throughput compared to the state-of-the-art approach.

cs.DC↗

arXiv2023

FastFold: Reducing AlphaFold Training Time from 11 Days to 67 Hours

Protein structure prediction helps to understand gene translation and protein function, which is of growing interest and importance in structural biology. The AlphaFold model, which used transformer architecture to achieve atomic-level accuracy in protein structure prediction, was a significant breakthrough. However, training and inference of the AlphaFold model are challenging due to its high computation and memory cost. In this work, we present FastFold, an efficient implementation of AlphaFold for both training and inference. We propose Dynamic Axial Parallelism and Duality Async Operations to improve the scaling efficiency of model parallelism. Besides, AutoChunk is proposed to reduce memory cost by over 80% during inference by automatically determining the chunk strategy. Experimental results show that FastFold reduces overall training time from 11 days to 67 hours and achieves 7.5X - 9.5X speedup for long-sequence inference. Furthermore, we scale FastFold to 512 GPUs and achieve an aggregate throughput of 6.02 PetaFLOP/s with 90.1% parallel efficiency.

cs.LG↗

arXiv2023

ATP: Adaptive Tensor Parallelism for Foundation Models

Foundation models have impressive performance and generalization capabilities across a wide range of applications. The increasing size of the models introduces great challenges for the training. Tensor parallelism is a critical technique that is currently used in almost all foundation model training and has a significant impact on overall training performance. However, current tensor parallelism in machine learning frameworks misses optimization opportunities in fitting various interconnection topologies. In this work, we present ATP, an adaptive tensor parallelism framework for foundation models, which can automatically select the optimal parallel strategy on different interconnections. We propose column- and row-first tensor parallelism based on 2D device meshes and construct a search space. Combined with the hierarchical communication matrix, ATP can identify the optimal strategy in the search space. We also propose chunk-based overlapping to reduce communication overhead. Our evaluations show ATP consistently outperforms the state-of-the-art approaches for various model sizes and interconnects, achieving end-to-end training performance improvements of up to 37-64% on specific interconnects. Based on our theoretical model, the communication overhead of ATP decreases with scaling, indicating a qualitative leap forward.

cs.DC↗

arXiv2021

tcFFT: Accelerating Half-Precision FFT through Tensor Cores

Fast Fourier Transform (FFT) is an essential tool in scientific and engineering computation. The increasing demand for mixed-precision FFT has made it possible to utilize half-precision floating-point (FP16) arithmetic for faster speed and energy saving. Specializing in lower precision, NVIDIA Tensor Cores can deliver extremely high computation performance. However, the fixed computation pattern makes it hard to utilize the computing power of Tensor Cores in FFT. Therefore, we developed tcFFT to accelerate FFT with Tensor Cores. Our tcFFT supports batched 1D and 2D FFT of various sizes and it exploits a set of optimizations to achieve high performance: 1) single-element manipulation on Tensor Core fragments to support special operations needed by FFT; 2) fine-grained data arrangement design to coordinate with the GPU memory access pattern. We evaluated our tcFFT and the NVIDIA cuFFT in various sizes and dimensions on NVIDIA V100 and A100 GPUs. The results show that our tcFFT can outperform cuFFT 1.29x-3.24x and 1.10x-3.03x on the two GPUs, respectively. Our tcFFT has a great potential for mixed-precision scientific applications.

cs.DC↗

arXiv2020

CUBE -- Towards an Optimal Scaling of Cosmological N-body Simulations

N-body simulations are essential tools in physical cosmology to understand the large-scale structure (LSS) formation of the Universe. Large-scale simulations with high resolution are important for exploring the substructure of universe and for determining fundamental physical parameters like neutrino mass. However, traditional particle-mesh (PM) based algorithms use considerable amounts of memory, which limits the scalability of simulations. Therefore, we designed a two-level PM algorithm CUBE towards optimal performance in memory consumption reduction. By using the fixed-point compression technique, CUBE reduces the memory consumption per N-body particle toward 6 bytes, an order of magnitude lower than the traditional PM-based algorithms. We scaled CUBE to 512 nodes (20,480 cores) on an Intel Cascade Lake based supercomputer with $\simeq$95\% weak-scaling efficiency. This scaling test was performed in "Cosmo-$π$" -- a cosmological LSS simulation using $\simeq$4.4 trillion particles, tracing the evolution of the universe over $\simeq$13.7 billion years. To our best knowledge, Cosmo-$π$ is the largest completed cosmological N-body simulation. We believe CUBE has a huge potential to scale on exascale supercomputers for larger simulations.

physics.comp-ph↗

arXiv2020

HMS-Net: Hierarchical Multi-scale Sparsity-invariant Network for Sparse Depth Completion

Dense depth cues are important and have wide applications in various computer vision tasks. In autonomous driving, LIDAR sensors are adopted to acquire depth measurements around the vehicle to perceive the surrounding environments. However, depth maps obtained by LIDAR are generally sparse because of its hardware limitation. The task of depth completion attracts increasing attention, which aims at generating a dense depth map from an input sparse depth map. To effectively utilize multi-scale features, we propose three novel sparsity-invariant operations, based on which, a sparsity-invariant multi-scale encoder-decoder network (HMS-Net) for handling sparse inputs and sparse feature maps is also proposed. Additional RGB features could be incorporated to further improve the depth completion performance. Our extensive experiments and component analysis on two public benchmarks, KITTI depth completion benchmark and NYU-depth-v2 dataset, demonstrate the effectiveness of the proposed approach. As of Aug. 12th, 2018, on KITTI depth completion leaderboard, our proposed model without RGB guidance ranks first among all peer-reviewed methods without using RGB information, and our model with RGB guidance ranks second among all RGB-guided methods.

cs.CV↗