GitHub

Ray.jl

Ray.Ray — Module
Ray

This package provides user-facing interface for Julia-on-Ray.

source
Ray.RayError — Type
RayError <: Exception

Abstract super type of all Ray exception types.

source
Ray.RaySystemError — Type
RaySystemError <: RayError

Indicates that Ray encountered a system error.

This exception is thrown when:

  • The raylet is killed.
  • Deserialization of a ObjectRef contains an unknown metadata error type.
source
Ray.RayTaskError — Type
RayTaskError <: RayError

Indicates that a Ray task threw an exception during execution.

If a Ray task throws an exception during execution, a RayTaskError is stored for the Ray task's output. When the object is retrieved, the contained exception is detected and thrown thereby propogating the exception to the Ray task caller.

source

Non-exported functions and types

Ray.GLOBAL_STATE_ACCESSOR — Constant
const GLOBAL_STATE_ACCESSOR::Ref{ray_jll.GlobalStateAccessor}

Global binding for GCS client interface to access global state information. Currently only used to get the next job ID.

This is set during init and used there to get the Job ID for the driver.

source
Ray.OutOfMemoryError — Type
OutOfMemoryError <: RayError

Indicates that the node is running out of memory and is close to full.

This exception is thrown when the node is low on memory and tasks or actors are being evicted to free up memory.

source
Base.isready — Method
Base.isready(obj_ref::ObjectRef)

Check whether obj_ref has a value that's ready to be retrieved.

source
Base.wait — Method
Base.wait(obj_ref::ObjectRef) -> Nothing

Block until isready(obj_ref).

source
Ray.get — Method
Ray.get(obj_ref::ObjectRef)

Retrieves the data associated with the object reference from the object store. This method is blocking until the data is available in the local object store.

If the task that generated the ObjectRef failed with a Julia exception, the captured exception will be thrown on get.

source
Ray.get_all_reference_counts — Method
get_all_reference_counts()

For testing/debugging purposes, returns a Dict{ray_jll.ObjectID,Tuple{Int,Int}} containing the reference counts for each object ID that the local raylet knows about. The first count is the "local reference" count, and the second is the count of submitted tasks depending on the object.

source
Ray.get_job_id — Method
get_job_id() -> UInt32

Get the current job ID for this worker or driver. Job ID is the id of your Ray drivers that create tasks.

source
Ray.get_task_id — Method
get_task_id() -> String

Get the current task ID for this worker in hex format.

source
Ray.put — Method
Ray.put(data) -> ObjectRef

Store data in the object store. Returns an object reference which can used to retrieve the data with Ray.get.

source