DeepSpeed
Microsoft’s Apache-2.0 ZeRO optimizer for multi-GPU, multi-node LLM training.
Fit.
Catalog snapshot
- Platforms
- Linux, Macos, Windows, Docker
- Operating systems
- Linux, macOS, Windows
- Ram floor
- 32 GB
- Gpu
- Required · 16 GB VRAM
- Docker
- Yes
- Setup
- Advanced
- Deploy as
- Local, Self-hosted
- Commercial use
- Allowed
The ZeRO-2/3 configs are the usual starting points; multi-GPU expected.
Project.
Fetched 1
About.
DeepSpeed is Microsoft’s Apache-2.0 training optimizer: ZeRO memory sharding, multi-node scaling, and inference kernels that let fixed hardware train far larger models. A CLI library for serious training runs, not casual tuning.
ZeRO sharding · multi-node training · inference kernel optimizations · parameter offload
