COMET: A Novel Memory-Efficient Deep Learning Training Framework by Using Error-Bounded Lossy Compression
Summary: COMET uses adaptive, error-bounded lossy compression to save activations in DNN training, enabling larger models under tight GPU memory. It analyzes gradient error propagation and adaptively bounds errors, achieving up to 13.5x memory reduction with minimal accuracy loss. (summarized by gpt-5-nano on Feb 09 2026)
Incoming Non-self Citations Over Time
Authors
- 1. Sian Jin
- 2. Chengming Zhang
- 3. Xintong Jiang
- 4. Yunhe Feng
- 5. Hui Guan
- 6. Guanpeng Li
- 7. Shuaiwen Leon Song
- 8. Dingwen Tao
Incoming Citations (Sorted by Pagerank)
Showing 3 of 3 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 8,786 | AWARE: Workload-aware, Redundancy-exploiting Linear Algebra | 2023 | SIGMOD | 4.4521262e-05 |
| 9,326 | BladeDISC: Optimizing Dynamic Shape Machine Learning Workloads via Compiler Approach | 2023 | SIGMOD | 4.3556432e-05 |
| 9,806 | The Image Calculator: 10x Faster Image-AI Inference by Replacing JPEG with Self-designing Storage Format | 2024 | SIGMOD | 4.2805224e-05 |
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 1 of 1 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 683 | Cerebro: A Data System for Optimized Deep Learning Model Selection | 2020 | VLDB | 0.00018195476 |
Previous
Page 1 / 1
Next