ConceptAI & ML
Zero partitioning
Memory-efficient data parallelism strategies that distribute model states across processors to enable scaling to larger models.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Data parallelism1Distributed training1
Lecturers
1Connections
2The lectures that define it or assume it, and the concepts next to it.