Corpus ID: 195316378

Placeto: Learning Generalizable Device Placement Algorithms for Distributed Machine Learning

@inproceedings{Addanki2019PlacetoLG,
  title={Placeto: Learning Generalizable Device Placement Algorithms for Distributed Machine Learning},
  author={Ravichandra Addanki and S. Venkatakrishnan and S. Gupta and Hongzi Mao and M. Alizadeh},
  booktitle={NeurIPS},
  year={2019}
}
We present Placeto, a reinforcement learning (RL) approach to efficiently find device placements for distributed neural network training. Unlike prior approaches that only find a device placement for a specific computation graph, Placeto can learn generalizable device placement policies that can be applied to any graph. We propose two key ideas in our approach: (1) we represent the policy as performing iterative placement improvements, rather than outputting a placement in one shot; (2) we use… Expand
21 Citations
GDP: Generalized Device Placement for Dataflow Graphs
  • 11
  • PDF
A Single-Shot Generalized Device Placement for Large Dataflow Graphs
Chip Placement with Deep Reinforcement Learning
  • 46
  • PDF
Graph Representation Matters in Device Placement
  • Highly Influenced
  • PDF
GDDR: GNN-based Data-Driven Routing
  • Highly Influenced
  • PDF
Geometric deep reinforcement learning for dynamic DAG scheduling
  • PDF
A Deep Reinforcement Learning Method for Solving Task Mapping Problems with Dynamic Traffic on Parallel Systems
...
1
2
3
...

References

SHOWING 1-10 OF 28 REFERENCES
A Hierarchical Model for Device Placement
  • 68
  • Highly Influential
  • PDF
Spotlight: Optimizing Device Placement for Training Deep Neural Networks
  • 34
  • PDF
Device Placement Optimization with Reinforcement Learning
  • 226
  • Highly Influential
  • PDF
REGAL: Transfer Learning For Fast Optimization of Computation Graphs
  • 16
  • PDF
Efficient Neural Architecture Search via Parameter Sharing
  • 1,178
  • Highly Influential
  • PDF
Beyond Data and Model Parallelism for Deep Neural Networks
  • 133
  • PDF
Rethinking the Inception Architecture for Computer Vision
  • 11,074
  • Highly Influential
  • PDF
Massively Parallel Methods for Deep Reinforcement Learning
  • 299
  • PDF
Relational inductive biases, deep learning, and graph networks
  • 1,023
  • PDF
Learning scheduling algorithms for data processing clusters
  • 139
  • PDF
...
1
2
3
...