Switchml

15 April 2021

Very excited to share that our work on SwitchML is now available. SwitchML shows that distributed ML training workloads are often network-bound, and in-network aggregation using a programmable switch can resolve this bottleneck.

We’ve released the code as open source through the P4 Consortium, so check that out – or the NSDI 21’ paper, presentation video, or explanation through cat photos