Polaris: Multi-Fidelity Design Space Exploration of Deep Learning Accelerators
Chirag Sakhuja, Charles Hong, Calvin Lin · 2024
Read Polaris: Multi-Fidelity Design Space Exploration of Deep Learning Accelerators by Chirag Sakhuja, Charles Hong, Calvin Lin (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-15548.txt.
Description
Designing a chip that runs neural networks means testing millions of layouts, and full tests are far too slow. Chirag Sakhuja, Charles Hong, and Calvin Lin present a tool pair for that search in a 2024 paper. Their model, Starlight, learns from cheap rough estimates to predict costly detailed results. Transfer learning bridges the gap between the two test levels.
Starlight gives thousands of predictions per second and matches detailed simulation closely. It also trains with far fewer samples than the prior best model. The search tool built on it, Polaris, matched rival designs in a fraction of the time. Given more hours, it found designs with much lower energy-delay scores. The work speeds the hunt for efficient AI hardware.
Catalog record
| CREATOR | Chirag Sakhuja, Charles Hong, Calvin Lin |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-15548