# FSDP **Entity class:** Distributed training technique **As of:** November 17, 2026 ## Definition **FSDP** is PyTorch's fully sharded data parallel approach for dividing model state across accelerator workers. This node belongs to the **distributed training** route of [[wiki/Distributed Compute Ecosystem|Distributed Compute Ecosystem]], which records the specific kind of work or network capacity involved. ## Relationships - **comparison:** [[wiki/ZeRO|ZeRO]]. - **mechanism route:** [[wiki/Distributed Compute Ecosystem|Distributed Compute Ecosystem]]. - **source article:** [[articles/Embedded Super-Spreader Events and Muse|Embedded Super-Spreader Events and Muse]]. ## Sources - [FSDP — project record or attributed account](https://docs.pytorch.org/docs/stable/fsdp).