MLsys 2021 [Proceeding]
-
Boveda: Building an On-Chip Deep Learning Memory Hierarchy Brick by Brick [paper]
- on-chip data compression for DNN
- not my expertise. welcome to leaving your comments.
-
Value Learning for Throughput Optimization of Deep Neural Networks [paper]
- model DNN optimization as a MDP (state + action）
- Exploring the state space efficeienctly: