Conference
EPLoN: Exploiting Efficient Parallelism with Selective Rematerialization for Lightning Attention on Ascend NPU
| Τίτλος: | EPLoN: Exploiting Efficient Parallelism with Selective Rematerialization for Lightning Attention on Ascend NPU |
|---|---|
| Συγγραφείς: | Bao, Han, Su, Zhenfeng, Setyaev, Alexander, Kamenev, Stanislav, Gneushev, Alexander, Zhao, Kang, Xiao, Junmin, Lin, Haoran, Bistrigova, Anastasiya, Buzykanov, Sergey, Tetin, Evgeny, Tan, Guangming, Liu, Boxiao, Zou, Xueyi, Dong, Zhenhua, Korikov, Constantine, Yu, Xianzhi, Hu, Zhongzhe |
| Πηγή: | Proceedings of the 40th ACM International Conference on Supercomputing. :1040-1052 |
| Διαθεσιμότητα: | http://dl.acm.org/doi/10.1145/3797905.3800540 |
| Βάση Δεδομένων: | ACM Full-Text Collection |
καταχωρήστε σχόλιο πρώτοι!