HetPipe - Enabling Large DNN Training on Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism

HetPipe - Enabling Large DNN Training on Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism

USENIX via YouTube Direct link

Conclusion

11 of 11

11 of 11

Conclusion

Class Central Classrooms beta

YouTube videos curated by Class Central.

Classroom Contents

HetPipe - Enabling Large DNN Training on Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism

Automatically move to the next video in the Classroom when playback concludes

  1. 1 Introduction
  2. 2 Motivation Background
  3. 3 Single Virtual Occur
  4. 4 Evaluation
  5. 5 Partitioning
  6. 6 Equal Distribution
  7. 7 Hybrid Policy
  8. 8 Parameter Placement Policy
  9. 9 Local Placement Policy
  10. 10 Convergence
  11. 11 Conclusion

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.