Evaluating supernets for neural architecture search

Gabriel Meyer-Lee, Nick Cheney

TL;DR: Making richer use of NAS benchmarks to evaluate trained supernets independently of the final architecture selected helps understand supernet training

Keywords: supernet, evaluation, open source

Broader Impact Statement: Yes

Paper Availability And License: Yes

Code Of Conduct: Yes

Cpu Hours: 0

Gpu Hours: 925

Tpu Hours: 0

Submission Number: 11