Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
SingleDeviceGraph + PartitionAnnotations -> DistributedExecutionPlan
Resolve and compile a distributed parallel execution strategy from minimal single-device model graphs with tensor partitioning annotations.
Problem it solves
Manual orchestration of hybrid parallel pipelines (data, tensor, pipeline) across hardware clusters is complex and hard to maintain.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.