Training a Trillion Parameters with Pipeline Parallelism

DeepSpeed includes new support for pipeline parallelism! DeepSpeed’s training engine provides hybrid 3D parallelism for training models with over a trillion parameters. In addition to scaling to the extreme, we have demonstrated that hybrid parallelism accelerates training on clusters with low-bandwidth network by up to 7x.

• For a brief overview and results including trillion-parameter capabilities, see our press release.
• To get started with pipeline parallel training in DeepSpeed, we recommend our tutorial.
• See our AlexNet example in DeepSpeedExamples.
• Read our API documentation on readthedocs.