Closed Dando18 closed 4 years ago
Add a distributed SGD optimizer to streamline the process of distributed training. This removes the necessity to write custom MPI/NCCL programs for distributed training.
Closed in 1f86e36c693a5329ef7700f694b33c9ff3873bb9
Add a distributed SGD optimizer to streamline the process of distributed training. This removes the necessity to write custom MPI/NCCL programs for distributed training.