Fetching the paper…

Improving Strong-Scaling of CNN Training by Exploiting Finer-Grained Parallelism · Around