4 citations · 11 across the 23 of their papers we have counts for
24 papers
Efficient Resource-Shiftable RL Post-training with Generative Rewards
Yanfeng Zhao, Haoyang Li, Sheng Lin +4
Reinforcement learning (RL) post-training with generative reward models (GenRM) has become a central recipe for improving large language models (LLMs), and it is also increasingly…
DFlow: Enabling Verifier Information Flow in Block Diffusion Speculative Decoding
Yaojie Zhang, Linfeng Zhang, Bin Cui +1
Block diffusion speculative decoding improves LLM inference efficiency by proposing a block of future tokens in parallel and verifying them with a single forward pass through the t…
QLPO: Quadrant-weighted Sampling for Length-aware Policy Optimization
Siwei Chen, Siqi Chen, Xupeng Miao +1
Recent large reasoning models often develop long chain-of-thought responses during reinforcement learning (RL), resulting in high inference latency and deployment cost. Existing me…
Adaptive Resource Management and Quality Control for Streaming Video Generation
Yifei Xia, Hao Yuan, Suhan Ling +5
Autoregressive diffusion transformers (AR-DiTs) recast video generation from an offline paradigm to a real-time streaming one: the model generates video one chunk at a time, making…
Harnessing Routing Foresight for Micro-step-level MoE load balancing in RL Post-training
Yuming Zhou, Haoyang Li, Sheng Lin +6
Mixture-of-Experts (MoE) and reinforcement learning (RL) post-training now dominate large language model (LLM) development, yet expert load imbalance remains a critical challenge.…
DARTS: Distribution-Aware Active Rollout Trajectory Shaping for Accelerating LLM Reinforcement Learning
Yujie Wang, Siwei Chen, Longzan Luo +4
Reinforcement Learning (RL) has become pivotal for improving model capabilities yet suffers from rollout efficiency bottlenecks due to the long-tail response length distribution. W…
LAER-MoE: Load-Adaptive Expert Re-layout for Efficient Mixture-of-Experts Training
Xinyi Liu, Yujie Wang, Fangcheng Fu +4
Expert parallelism is vital for effectively training Mixture-of-Experts (MoE) models, enabling different devices to host distinct experts, with each device processing different inp…
Efficient Multi-round LLM Inference over Disaggregated Serving
Wenhao He, Youhe Jiang, Penghao Zhao +4
With the rapid evolution of Large Language Models (LLMs), multi-round workflows, such as autonomous agents and iterative retrieval, have become increasingly prevalent. However, thi…
StaleFlow: Staleness-Aware Data Management for Mitigating Data Skewness in Fully Disaggregated RL Post-Training
Haoyang Li, Sheng Lin, Fangcheng Fu +6
Reinforcement learning (RL) post-training has become pivotal for enhancing the capabilities of modern large models. A recent trend is to develop RL systems with a fully disaggregat…
TridentServe: A Stage-level Serving System for Diffusion Pipelines
Yifei Xia, Fangcheng Fu, Hao Yuan +6
Diffusion pipelines, renowned for their powerful visual generation capabilities, have seen widespread adoption in generative vision tasks (e.g., text-to-image/video). These pipelin…
Concertina: Data-Centric Adaptive Pipeline Parallelism for Efficient Heterogeneous Long-Context LLM Training
Shiju Wang, Yujie Wang, Ao Sun +6
Long context training is crucial for extending LLM context windows. Existing schemes, such as sequence parallelism, incur substantial communication overhead. Pipeline parallelism (…
SALE : Low-bit Estimation for Efficient Sparse Attention in Long-context LLM Prefilling
Xiaodong Ji, Hailin Zhang, Fangcheng Fu +1
Many advanced Large Language Model (LLM) applications require long-context processing, but the self-attention module becomes a bottleneck during the prefilling stage of inference d…
Thinking Short and Right Over Thinking Long: Serving LLM Reasoning Efficiently and Accurately
Yuhang Wang, Youhe Jiang, Bin Cui +1
Recent advances in test-time scaling suggest that Large Language Models (LLMs) can gain better capabilities by generating Chain-of-Thought reasoning (analogous to human thinking) t…
Galvatron: An Automatic Distributed System for Efficient Foundation Model Training
Xinyi Liu, Yujie Wang, Shenhan Zhu +4
Galvatron is a distributed system for efficiently training large-scale Foundation Models. It overcomes the complexities of selecting optimal parallelism strategies by automatically…
Hetu v2: A General and Scalable Deep Learning System with Hierarchical and Heterogeneous Single Program Multiple Data Annotations
Haoyang Li, Fangcheng Fu, Hao Ge +6
The Single-Program Multiple-Data (SPMD) paradigm provides a unified abstraction to annotate various parallel dimensions in distributed deep learning (DL) training. With SPMD, users…
ByteScale: Efficient Scaling of LLM Training with a 2048K Context Length on More Than 12,000 GPUs
Hao Ge, Junda Feng, Qi Huang +6
Scaling long-context ability is essential for Large Language Models (LLMs). To amortize the memory consumption across multiple devices in long-context training, inter-data partitio…
Training-free and Adaptive Sparse Attention for Efficient Long Video Generation
Yifei Xia, Suhan Ling, Fangcheng Fu +4
Generating high-fidelity long videos with Diffusion Transformers (DiTs) is often hindered by significant latency, primarily due to the computational demands of attention mechanisms…
ThunderServe: High-performance and Cost-efficient LLM Serving in Cloud Environments
Youhe Jiang, Fangcheng Fu, Xiaozhe Yao +4
Recent developments in large language models (LLMs) have demonstrated their remarkable proficiency in a range of tasks. Compared to in-house homogeneous GPU clusters, deploying LLM…
Demystifying Cost-Efficiency in LLM Serving over Heterogeneous GPUs
Youhe Jiang, Fangcheng Fu, Xiaozhe Yao +6
Recent advancements in Large Language Models (LLMs) have led to increasingly diverse requests, accompanied with varying resource (compute and memory) demands to serve them. However…
Hydraulis: Balancing Large Transformer Model Training via Co-designing Parallel Strategies and Data Assignment
Haoyang Li, Fangcheng Fu, Sheng Lin +8
To optimize large Transformer model training, both efficient parallel computing and advanced data management are indispensable. However, current methods often assume a stable and u…
FlexSP: Accelerating Large Language Model Training via Flexible Sequence Parallelism
Yujie Wang, Shiju Wang, Shenhan Zhu +7
Extending the context length (i.e., the maximum supported sequence length) of LLMs is of paramount significance. To facilitate long context training of LLMs, sequence parallelism h…
LSH-MoE: Communication-efficient MoE Training via Locality-Sensitive Hashing
Xiaonan Nie, Qibin Liu, Fangcheng Fu +6
Larger transformer models always perform better on various tasks but require more costs to scale up the model size. To efficiently enlarge models, the mixture-of-experts (MoE) arch…
Malleus: Straggler-Resilient Hybrid Parallel Training of Large-scale Models via Malleable Data and Model Parallelization
Haoyang Li, Fangcheng Fu, Hao Ge +7
As the scale of models and training data continues to grow, there is an expanding reliance on more GPUs to train large-scale models, which inevitably increases the likelihood of en…
Spindle: Efficient Distributed Training of Multi-Task Large Models via Wavefront Scheduling
Yujie Wang, Shenhan Zhu, Fangcheng Fu +6
Recent foundation models are capable of handling multiple tasks and multiple data modalities with the unified base model structure and several specialized model components. However…