124 citations
- Intel (United States)US7 papers
- Georgia Institute of TechnologyUS5 papers
- ETH ZurichCH2 papers
- Indian Institute of Technology KharagpurIN2 papers
- Indian Institute of Technology MadrasIN2 papers
- Max Planck Institute for AstrophysicsDE2 papers
- Menlo SchoolUS2 papers
- Robert Bosch (India)IN2 papers
- Arizona State UniversityUS1 paper
- Astronomical Observatory of RomeIT1 paper
- Atlanta Technical CollegeUS1 paper
- California Southern UniversityUS1 paper
4 papers · 1 filter
Themis: A Network Bandwidth-Aware Collective Scheduling Policy for Distributed Training of DL Models
Saeed Rashidi, William Won, Sudarshan Srinivasan +2
Distributed training is a solution to reduce DNN training time by splitting the task across multiple NPUs (e.g., GPU/TPU). However, distributed training adds communication overhead…
Pythia: A Customizable Hardware Prefetching Framework Using Online Reinforcement Learning
Rahul Bera, Konstantinos Kanellopoulos, Anant V. Nori +3
Past research has proposed numerous hardware prefetching techniques, most of which rely on exploiting one specific type of program context information (e.g., program counter, cache…
LIBRA: Enabling Workload-aware Multi-dimensional Network Topology Optimization for Distributed Training of Large AI Models
William Won, Saeed Rashidi, Sudarshan Srinivasan +1
As model sizes in machine learning continue to scale, distributed training is necessary to accommodate model weights within each device and to reduce training time. However, this c…
Extending Sparse Tensor Accelerators to Support Multiple Compression Formats
Eric Qin, Geonhwa Jeong, William Won +7
Sparsity, which occurs in both scientific applications and Deep Learning (DL) models, has been a key target of optimization within recent ASIC accelerators due to the potential mem…