TOOLORA / Local reading edition
DeepSpeed
Optimize distributed deep-learning training, memory and inference.
Overview
DeepSpeed is an optimization library for deep-learning training and inference. Developers configure capabilities such as ZeRO memory optimization, mixed precision, distributed communication and parallel inference. Tutorials, configuration references and profiling tools support engineering workflows for training and running models.
- Company country / region
- United States · Published company address
Official sources and references
- Official websitehttps://www.deepspeed.ai/
- Documentationhttps://deepspeed.readthedocs.io/
- Documentationhttps://www.deepspeed.ai/docs/config-json/
Descriptions summarize public information; consult official websites for features, pricing and regional availability.