Distributed training with MindSpore can solve the recent rise in bottlenecks on ultra-large-scale networks due to huge increases in data & parameters.