REVIEW 21 cited by
Striped Attention: Faster Ring Attention for Causal Transformers
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
Striped Attention: Faster Ring Attention for Causal Transformers
read the original abstract
To help address the growing demand for ever-longer sequence lengths in transformer models, Liu et al. recently proposed Ring Attention, an exact attention algorithm capable of overcoming per-device memory bottle- necks by distributing self-attention across multiple devices. In this paper, we study the performance characteristics of Ring Attention in the important special case of causal transformer models, and identify a key workload imbal- ance due to triangular structure of causal attention computations. We propose a simple extension to Ring Attention, which we call Striped Attention to fix this imbalance. Instead of devices having contiguous subsequences, each device has a subset of tokens distributed uniformly throughout the sequence, which we demonstrate leads to more even workloads. In experiments running Striped Attention on A100 GPUs and TPUv4s, we are able to achieve up to 1.45x end-to-end throughput improvements over the original Ring Attention algorithm on causal transformer training at a sequence length of 256k. Furthermore, on 16 TPUv4 chips, we were able to achieve 1.65x speedups at sequence lengths of 786k. We release the code for our experiments as open source
Forward citations
Cited by 21 Pith papers
-
HCMS: Head-Chunked Multi-Stream Pipeline for Communication-Computation Overlap in Long-Sequence Parallel Attention
HCMS partitions multi-head attention into chunks and pipelines them across dual CUDA streams to overlap communication and computation, delivering 10-17.5% speedup over Ulysses for 31K-56K token sequences.
-
FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision
FlashAttention-3 achieves 1.5-2x speedup on H100 GPUs for attention, reaching 740 TFLOPs/s (75% utilization) in FP16 and near 1.2 PFLOPs/s in FP8 while cutting numerical error by 2.6x versus baseline FP8 attention.
-
Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
Transformers and SSMs are unified through structured state space duality, producing a 2-8X faster Mamba-2 model that remains competitive with Transformers.
-
From Tensor Buffer to Distributed Memory Hierarchy: A Survey of KV Cache Management for LLM Serving
KV-cache serving systems concentrate into five archetypes under a four-axis taxonomy, with ownership explaining residual distributed design variance and seven measurement gaps blocking next steps.
-
Libra: Taming Attention Workload Skew in Long-Context LLM Training with Bounded Sequence Pool
Libra bounds attention load-balancing to fixed-size sequence pools and combines variance-reduced sequence placement with tiled attention pooling, achieving up to 2.54x end-to-end throughput over Ulysses on 256K- and 1...
-
A Training-Memory Regression in MLA Sequence Parallelism: Why Megatron-Core Forbids Absorption, and LAGA -- a Communication-Efficient Fix
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.
-
ASAP: A Disaggregated and Asynchronous Inference System for MoE Prefill
ASAP is a disaggregated asynchronous inference system for the prefill phase of MoE models that removes DP-EP synchronization barriers and reports 90% higher SLO-compliant throughput than synchronous baselines.
-
Achieving Cloud-Grade SLOs for Local Mixture-of-Experts Inference through CPU-GPU Hybrid Design
A CPU-GPU hybrid design with stream-loading prefill, expert parallelism, and disaggregation achieves cloud SLOs for local MoE inference on dual-socket CPUs and consumer GPUs.
-
HexiSeq: Accommodating Long Context Training of LLMs over Heterogeneous Hardware
HexiSeq optimizes sequence and head partitioning across mixed GPUs to improve long-context LLM training throughput by up to 1.72x in simulations.
-
CoCoDiff: Optimizing Collective Communications for Distributed Diffusion Transformer Inference Under Ulysses Sequence Parallelism
CoCoDiff achieves 3.6x average and 8.4x peak speedup for distributed DiT inference on up to 96 GPU tiles via tile-aware all-to-all, V-first scheduling, and selective V communication.
-
DASH: Deterministic Attention Scheduling for High-throughput Reproducible LLM Training
DASH speeds up deterministic attention backprop by reversed and shift-based tile schedules, giving up to 1.28× throughput and ~5% end-to-end gains while keeping bitwise reproducibility.
-
MTraining: Distributed Dynamic Sparse Attention for Efficient Ultra-Long Context Training
MTraining scales LLM training to 512K-token contexts on 32 A100 GPUs by integrating dynamic sparse training patterns with balanced and hierarchical sparse ring attention, achieving up to 6x throughput gains without ac...
-
InfiniPipe: Elastic Pipeline Parallelism for Efficient Variable-Length Long-Context LLM Training
InfiniPipe proposes elastic pipeline parallelism and stage-aware chunk-level adaptive checkpointing to achieve 1.69x speedup over state-of-the-art for variable-length long-context LLM training.
-
Gated Linear Attention Transformers with Hardware-Efficient Training
Gated linear attention Transformers achieve competitive language modeling results with linear-time inference, superior length generalization, and higher training throughput than Mamba.
-
QLPO: Quadrant-weighted Sampling for Length-aware Policy Optimization
QLPO resamples GRPO training groups to favor short correct and long incorrect responses, cutting reasoning length substantially while keeping accuracy roughly unchanged.
-
Arachne: Orchestrating Cascades for Efficient Text-to-Video Model Training
Arachne orchestrates cascades for distributed T2V training and reports up to 65% lower iteration time with improving gains at larger scales compared to static bucketing approaches.
-
Kaczmarz Linear Attention
Kaczmarz Linear Attention replaces the empirical coefficient in Gated DeltaNet with a key-norm-normalized step size derived from the online regression objective, yielding lower perplexity and better needle-in-haystack...
-
Towards Distributed Inference of LLMs on a P2P Network
A decentralized prefix-cache-aware routing scheme for P2P LLM serving improves simulated latency under low-delay skewed workloads but is limited by network latency and hotspots.
-
Efficient Scaling of LLM Training with Flexible Context Parallelism
Regrouping context-parallel workers per micro-batch with non-power-of-two group sizes, chosen by a fitted cost model plus dynamic programming, is claimed to speed up multimodal LLM training by up to 1.36x (1.46x per t...
-
World Model on Million-Length Video And Language With Blockwise RingAttention
Presents open-source 7B models for million-token video and language understanding via Blockwise RingAttention, setting new benchmarks in retrieval and long video tasks.
-
Towards Efficient Large Vision-Language Models: A Comprehensive Survey on Inference Strategies
The paper surveys and taxonomizes inference optimization methods for large vision-language models across four categories while noting limitations and open problems.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.