本文介绍了11月更新的几项关键技术，包括GPU内存管理的BuddyCompression、Capuchin、DeepUM、SwapAdvisor和SuperNeurons，以及利用SSD的FlashNeuron和内存footprintreduction的Tempo，这些都致力于提升深度学习和HPC工作的内存效率。
11月week1更新
1.Buddy Compression: Enabling Larger Memory for Deep Learning and HPC Workloads on GPUs（11月week1更新）
2.Capuchin: Tensor-based GPU Memory Management for Deep Learning（11月week1更新）
3.DeepUM: Tensor Migration and Prefetching in Unified Memory（11月week1更新）
4.SwapAdvisor: Pushing Deep Learning Beyond the GPU Memory Limit via Smart Swapping（11月week1更新）
5.SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks（11月week1更新）
6.FlashNeuron: SSD-Enabled Large-Batch Training of Very Deep Neural Networks（11月week1更新）
7.Tempo: Accelerating Transformer-Based Model Training through Memory Footprint Reduction（11月week1更新）
扫一扫
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
