Our paper “NASPipe: High Performance and Reproducible Pipeline Parallel Supernet Training via Causal Synchronous Parallel” was accepted by ASPLOS ‘22.