Zobrazeno 1 - 2
of 2
pro vyhledávání: '"Ouyang, Anne"'
Autor:
Ouyang, Anne
The state of the art results in natural language processing tasks have been obtained by scaling up transformer-based machine learning models, which can have more than a hundred billion parameters. Training and deploying these models can be difficult
Externí odkaz:
https://hdl.handle.net/1721.1/151543
Autor:
Kaler, Tim, Stathas, Nickolas, Ouyang, Anne, Iliopoulos, Alexandros-Stavros, Schardl, Tao B., Leiserson, Charles E., Chen, Jie
Improving the training and inference performance of graph neural networks (GNNs) is faced with a challenge uncommon in general neural networks: creating mini-batches requires a lot of computation and data movement due to the exponential growth of mul
Externí odkaz:
http://arxiv.org/abs/2110.08450