Abstract | ||
---|---|---|
This paper introduces a new technique for mapping Deep Recurrent Neural Networks (RNN) efficiently onto GPUs. We show how it is possible to achieve substantially higher computational throughput at low mini-batch sizes than direct implementations of RNNs based on matrix multiplications. The key to our approach is the use of persistent computational kernels that exploit the GPUu0027s inverted memory hierarchy to reuse network weights over multiple timesteps. Our initial implementation sustains 2.8 TFLOP/s at a mini-batch size of 4 on an NVIDIA TitanX GPU. This provides a 16× reduction in activation memory footprint, enables model training with 12× more parameters on the same hardware, allows us to strongly scale RNN training to 128 GPUs, and allows us to efficiently explore end-to-end speech recognition models with over 100 layers. |
Year | Venue | Field |
---|---|---|
2016 | ICML | Memory hierarchy,CUDA,Computer science,Parallel computing,Recurrent neural network,Artificial intelligence,Throughput,Deep learning,Memory footprint,Artificial neural network,Matrix multiplication,Machine learning |
DocType | Citations | PageRank |
Conference | 12 | 0.73 |
References | Authors | |
19 | 9 |
Name | Order | Citations | PageRank |
---|---|---|---|
Gregory Frederick Diamos | 1 | 1117 | 51.07 |
Shubho Sengupta | 2 | 505 | 19.84 |
Bryan C. Catanzaro | 3 | 1191 | 75.56 |
mike chrzanowski | 4 | 309 | 12.21 |
Adam Coates | 5 | 2493 | 160.95 |
Erich Elsen | 6 | 551 | 29.33 |
Jesse H. Engel | 7 | 326 | 20.21 |
Awni Y. Hannun | 8 | 517 | 27.54 |
Sanjeev Satheesh | 9 | 5591 | 233.55 |