ray-project / ray

Ray is a unified framework for scaling AI and Python applications. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.

Home Page:https://ray.io

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[core][experimental] Avoid serialization for data passed between two tasks on the same actor

stephanie-wang opened this issue · comments

Description

#44259 supports multiple tasks on the same actor in one DAG. However, if the same actor passes data from one task to the next, it currently still requires a shared memory channel, which means we need to copy and serialize/deserialize. We should add a LocalChannel type that just buffers the Python value to avoid this.

Use case

No response