Responding to the "datacenter tax" and "killer microseconds" problems fo...
The advent of switches with programmable dataplanes has enabled the rapi...
Large ML models and datasets have necessitated the use of multi-GPU syst...
ML workloads are becoming increasingly popular in the cloud. Good cloud
...
We present bundled references, a new building block to provide lineariza...
Collective communication algorithms are an important component of distri...
Training complex machine learning models in parallel is an increasingly
...
Distributed deep neural network (DDNN) training constitutes an increasin...
Most work in the deep learning systems community has focused on faster
i...