Octo - INT8 Training with Loss-aware Compensation and Backward Quantization for Tiny On-device Learning

Octo - INT8 Training with Loss-aware Compensation and Backward Quantization for Tiny On-device Learning

USENIX via YouTube Direct link

Why We Need Quantization?

6 of 18

6 of 18

Why We Need Quantization?

Class Central Classrooms beta

YouTube videos curated by Class Central.

Classroom Contents

Octo - INT8 Training with Loss-aware Compensation and Backward Quantization for Tiny On-device Learning

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

  1. 1 Intro
  2. 2 Rise of On-device Learning
  3. 3 Common Compression Methods
  4. 4 The Workflow of DNN Training
  5. 5 Bridge the Gap: Data Quantization
  6. 6 Why We Need Quantization?
  7. 7 Potential Gains
  8. 8 Co-design of Network and Training Engine
  9. 9 Our System: Octo
  10. 10 Loss-aware Compensation
  11. 11 Backward Quantization
  12. 12 Evaluation Setup
  13. 13 Convergence Results
  14. 14 Ablation Study: Impact of LAC and PRC
  15. 15 Image Processing Throughput
  16. 16 Deep Insight of Feature Distribution Visualization of intermediate Feature Distribution
  17. 17 System Overhead
  18. 18 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.