203 citations · 593 across the 26 of their papers we have counts for
13 papers · 1 filter
ParamSpMM: Adaptive and Efficient Sparse Matrix-Matrix Multiplication on GPUs for GNNs
Lixing Zhang, Guanhua Ye, Hongzheng Li +2
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.…
TRANSOM: An Efficient Fault-Tolerant System for Training LLMs
Baodong Wu, Lei Xia, Qingping Li +6
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 w…
AutoDDL: Automatic Distributed Deep Learning with Near-Optimal Bandwidth Cost
Jinfan Chen, Shigang Li, Ran Gun +2
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…
HammingMesh: A Network Topology for Large-Scale Deep Learning
Torsten Hoefler, Tommaso Bonato, Daniele De Sensi +7
Numerous microarchitectural optimizations unlocked tremendous processing power for deep neural networks that in turn fueled the AI revolution. With the exhaustion of such optimizat…
Efficient Quantized Sparse Matrix Operations on Tensor Cores
Shigang Li, Kazuki Osawa, Torsten Hoefler
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 spa…
Near-Optimal Sparse Allreduce for Distributed Deep Learning
Shigang Li, Torsten Hoefler
Communication overhead is one of the major obstacles to train large deep learning models at scale. Gradient sparsification is a promising technique to reduce the communication volu…
Chimera: Efficiently Training Large-Scale Neural Networks with Bidirectional Pipelines
Shigang Li, Torsten Hoefler
Training large deep learning models at scale is very challenging. This paper proposes Chimera, a novel pipeline parallelism scheme which combines bidirectional pipelines for effici…
Flare: Flexible In-Network Allreduce
Daniele De Sensi, Salvatore Di Girolamo, Saleh Ashkboos +2
The allreduce operation is one of the most commonly used communication routines in distributed applications. To improve its bandwidth and to reduce network traffic, this operation…
AGCM-3DLF: Accelerating Atmospheric General Circulation Model via 3D Parallelization and Leap-Format
Hang Cao, Liang Yuan, He Zhang +1
The Atmospheric General Circulation Model (AGCM) has been an important research tool in the study of climate change for decades. As the demand for high-resolution simulation is bec…
Breaking (Global) Barriers in Parallel Stochastic Optimization with Wait-Avoiding Group Averaging
Shigang Li, Tal Ben-Nun, Giorgi Nadiradze +4
Deep learning at scale is dominated by communication time. Distributing samples across nodes usually yields the best performance, but poses scaling challenges due to global informa…
The Scalability for Parallel Machine Learning Training Algorithm: Dataset Matters
Daning Cheng, Hanping Zhang, Fen Xia +2
To gain a better performance, many researchers put more computing resource into an application. However, in the AI area, there is still a lack of a successful large-scale machine l…
Taming Unbalanced Training Workloads in Deep Learning with Partial Collective Operations
Shigang Li, Tal Ben-Nun, Salvatore Di Girolamo +2
Load imbalance pervasively exists in distributed deep learning training systems, either caused by the inherent imbalance in learned tasks or by the system itself. Traditional synch…
Asynchronous COMID: the theoretic basis for transmitted data sparsification tricks on Parameter Server
Daning Cheng, Shigang Li, Yunquan Zhang
Asynchronous FTRL-proximal and L2 norm done at server are two widely used tricks in Parameters Server which is an implement of delayed SGD. Their commonness is leaving parts of upd…