171 citations · 200 across the 3 of their papers we have counts for
6 papers · 1 filter
MegaBlocks: Efficient Sparse Training with Mixture-of-Experts
Trevor Gale, Deepak Narayanan, Cliff Young +1
We present MegaBlocks, a system for efficient Mixture-of-Experts (MoE) training on GPUs. Our system is motivated by the limitations of current frameworks, which restrict the dynami…
Exploring the limits of Concurrency in ML Training on Google TPUs
Sameer Kumar, James Bradbury, Cliff Young +16
Recent results in language understanding using neural networks have required training hardware of unprecedentedscale, with thousands of chips cooperating on a single training run.…
Sparse GPU Kernels for Deep Learning
Trevor Gale, Matei Zaharia, Cliff Young +1
Scientific workloads have traditionally exploited high levels of sparsity to accelerate computation and reduce memory requirements. While deep neural networks can be made sparse, a…
Bit-Parallel Vector Composability for Neural Acceleration
Soroush Ghodrati, Hardik Sharma, Cliff Young +2
Conventional neural accelerators rely on isolated self-sufficient functional units that perform an atomic operation while communicating the results through an operand delivery-aggr…
MLPerf Training Benchmark
Peter Mattson, Christine Cheng, Cody Coleman +34
Machine learning (ML) needs industry-standard performance benchmarks to support design and competitive evaluation of the many emerging software and hardware solutions for ML. But M…
Mesh-TensorFlow: Deep Learning for Supercomputers
Noam Shazeer, Youlong Cheng, Niki Parmar +9
Batch-splitting (data-parallelism) is the dominant distributed Deep Neural Network (DNN) training strategy, due to its universal applicability and its amenability to Single-Program…