Pith. sign in

REVIEW 23 cited by

USP: A Unified Sequence Parallelism Approach for Long Context Generative AI

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2405.07719 v5 pith:K65KSBEJ submitted 2024-05-13 cs.LG cs.AI

classification cs.LGcs.AI
keywords parallelismsequenceapproachgenerativemodelunifiedachievedacross
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Sequence parallelism (SP), which divides the sequence dimension of input tensors across multiple computational devices, is becoming key to unlocking the long-context capabilities of generative AI models. This paper investigates the state-of-the-art SP approaches, i.e. DeepSpeed-Ulysses and Ring-Attention, and proposes a unified SP approach, which is more robust to transformer model architectures and network hardware topology. This paper compares the communication and memory cost of SP and existing parallelism, including data/tensor/zero/pipeline parallelism, and discusses the best practices for designing hybrid 4D parallelism involving SP. We achieved 47% MFU on two 8xA800 nodes using SP for the LLAMA3-8B model training using sequence length 208K. Our code is publicly available at https://github.com/feifeibear/long-context-attention.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 23 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Zellige: Moldable Sequence Placement for Mixed Image-Video DiT Training

    cs.DC 2026-08 conditional novelty 7.0 of 10

    Zellige assigns each training sequence its own parallelism configuration and GPU ranks, outperforming prior disjoint-group placement systems by 1.12 to 1.54x in end-to-end DiT training.

  2. X-Stage: An Overlooked Pipeline Stage for Communication-Computation Overlap in DiT Inference

    cs.DC 2026-07 conditional novelty 7.0 of 10

    X-Stage is a measurable post-issue window for GPU remote stores; a Burst-Gap model predicts backpressure and guides scheduling, yielding up to 1.62x MoE and 1.43x attention speedups.

  3. LiveAnimate: Stable Long-Form Streaming Human Animation in Real-Time

    cs.CV 2026-08 conditional novelty 6.0 of 10

    LiveAnimate is a 14-billion-parameter video diffusion system that generates stable, identity-consistent human animation at about 20 frames per second from a streaming pose input.

  4. StateFlow: Sequence Pipeline Parallelism for Long-Context Modeling with Linear Recurrence

    cs.DC 2026-08 conditional novelty 6.0 of 10

    StateFlow is a sequence pipeline parallelism system that propagates boundary states across sequence chunks to reduce activation memory and pipeline bubbles when training linear-recurrence and hybrid models.

  5. A Training-Memory Regression in MLA Sequence Parallelism: Why Megatron-Core Forbids Absorption, and LAGA -- a Communication-Efficient Fix

    cs.DC 2026-07 conditional novelty 6.0 of 10

    LAGA replaces MLA training's per-head K/V all-to-all with a latent all-gather and local up-projection, matching explicit-form memory while cutting collective communication ~1.98x.

  6. Audio-Visual Flamingo: Open Audio-Visual Intelligence for Long and Complex Videos

    eess.AS 2026-07 conditional novelty 6.0 of 10

    AV-Flamingo is an open audio-visual language model that outperforms similarly sized open models on many long-video understanding benchmarks, using a new 7M-question dataset and timestamped reasoning.

  7. DiTango: Cost-Effective Parallel Diffusion Generation with Selective Attention State Reuse

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Selective reuse of composed attention states across denoising steps lets DiTango skip both remote KV communication and attention compute for low-contribution sequence partitions, cutting multi-GPU diffusion latency by...

  8. Wan-Dancer: A Hierarchical Framework for Minute-scale Coherent Music-to-Dance Generation

    cs.CV 2026-07 conditional novelty 6.0 of 10

    A hierarchical global-keyframe then local-refinement diffusion pipeline produces stable 720p/30fps music-to-dance videos longer than one minute across five genres.

  9. LSRM: High-Fidelity Object-Centric Reconstruction via Scaled Context Windows

    cs.CV 2026-04 conditional novelty 6.0 of 10

    Scaling transformer context with sparse attention and 3D-aware block routing improves feed-forward 3D reconstruction and inverse rendering, closing much of the quality gap with dense-view optimization.

  10. Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking

    cs.LG 2026-02 conditional novelty 6.0 of 10

    UPipe chunks attention by head so QKV and all-to-all buffers scale with a small tunable chunk size rather than head count, enabling 5M-token Llama3-8B training on one 8×H100 node with throughput close to Ulysses.

  11. SwiftFusion: Scalable Sequence Parallelism for Distributed Inference of Diffusion Transformers on GPUs

    cs.DC 2026-01 conditional novelty 6.0 of 10

    StreamFusion places Ulysses-style all-to-all communication between GPU machines and ring-style communication inside machines, overlaps the inter-machine transfers with attention computation, and reports 1.35x average ...

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

    cs.LG 2025-07 conditional novelty 6.0 of 10

    HelixPipe repartitions transformer layers so attention runs in parallel across pipeline stages, cutting idle time and memory imbalance, and reports a 26% speedup for a 7B model on 128k-length sequences.

  13. Arctic Long Sequence Training: Scalable And Efficient Training For Multi-Million Token Sequences

    cs.LG 2025-06 conditional novelty 6.0 of 10

    ALST combines tiled loss and MLP compute, Ulysses-style sequence parallelism for Hugging Face GQA and MQA models, and activation-checkpoint CPU offload to train Llama-8B at up to 500K tokens on one H100 and 15M tokens...

  14. SRDiffusion: Accelerate Video Diffusion Inference via Sketching-Rendering Cooperation

    cs.GR 2025-05 conditional novelty 6.0 of 10

    SRDiffusion accelerates video diffusion by switching from a large model to a smaller sibling model after early high-noise steps, using an adaptive threshold for the switch.

  15. Domain Adaptation of Foundation LLMs for e-Commerce

    cs.CL 2025-01 conditional novelty 6.0 of 10

    Continued pretraining Llama 3.1 on 1 trillion e-commerce tokens produces e-Llama 8B/70B models that improve in-house e-commerce benchmarks by roughly 25-30% while retaining most general-domain accuracy.

  16. Visko Orbis 1.0: A Live Model for Real-Time Interactive Long Video Generation

    cs.CV 2026-07 conditional novelty 5.0 of 10

    A chunk-wise streaming video model with bounded multi-scale memory and streaming 4K upscaling reports real-time interactive long-form generation and top Arena preference/stability scores.

  17. MoWorld: A Flash World Model

    cs.CV 2026-07 conditional novelty 5.0 of 10

    MoWorld achieves up to 50 FPS real-time interactive world simulation on NPUs by combining a 3D-native data engine, curriculum cross-frame pretraining, autoregressive distillation, and mixed-precision parallel inference.

  18. Seeing Voices: Generating A-Roll Video from Audio with Mirage

    cs.CV 2025-06 reject novelty 5.0 of 10

    Mirage generates photorealistic A-roll videos of people speaking directly from audio, using only joint self-attention over audio, text, and video tokens.

  19. SkyReels-Audio: Omni Audio-Conditioned Talking Portraits in Video Diffusion Transformers

    cs.CV 2025-06 conditional novelty 5.0 of 10

    An audio-conditioned video diffusion transformer that animates portraits from image, video, text, and audio inputs with a sliding-window fusion for long videos.

  20. Wan-S2V: Audio-Driven Cinematic Video Generation

    cs.CV 2025-08 reject novelty 4.0 of 10

    Wan-S2V is an audio-driven video generator built on Wan, claiming better cinematic character animation than prior systems, though the evaluation is limited.

  21. TokenRing: An Efficient Parallelism Framework for Infinite-Context LLMs via Bidirectional Communication

    cs.DC 2024-12 conditional novelty 4.0 of 10

    TokenRing claims faster long-sequence attention on multiple GPUs by overlapping forward query transfers with backward partial-output transfers, supported by one 4-GPU profiling comparison.

  22. Next Token Prediction Towards Multimodal Intelligence: A Comprehensive Survey

    cs.CL 2024-12 conditional novelty 4.0 of 10

    A comprehensive survey that frames multimodal understanding and generation as next token prediction and proposes a five-part taxonomy.

  23. Efficient Diffusion Models: A Survey

    cs.LG 2025-02 conditional novelty 2.0 of 10

    The paper organizes research on efficient diffusion models into a taxonomy spanning algorithms, systems, and frameworks, and provides a curated reference list.

Pith tools