wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-06-27 19:16:19 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Doc] Document Actor.options API (#6099)

Browse Source 
* Document Actor.options API

* Undocument _remote
This commit is contained in:
Simon Mo
2019-11-06 23:12:23 -08:00
committed by GitHub
parent 9820c10a09
commit fcb6bdbc39
1 changed files with 10 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
doc/source/advanced.rst
+10 -9
View File
@@ -6,7 +6,7 @@ This page will cover some more advanced examples of using Ray's flexible program
Dynamic Remote Parameters
-------------------------
You can dynamically adjust resource requirements or return values of ``ray.remote`` during execution with ``._remote``.
You can dynamically adjust resource requirements or return values of ``ray.remote`` during execution with ``.options``.
For example, here we instantiate many copies of the same actor with varying resource requirements. Note that to create these actors successfully, Ray will need to be started with sufficient CPU resources and the relevant custom resources:
@@ -21,9 +21,9 @@ For example, here we instantiate many copies of the same actor with varying reso
self.value += 1
return self.value
a1 = Counter._remote(num_cpus=1, resources={"Custom1": 1})
a2 = Counter._remote(num_cpus=2, resources={"Custom2": 1})
a3 = Counter._remote(num_cpus=3, resources={"Custom3": 1})
a1 = Counter.options(num_cpus=1, resources={"Custom1": 1}).remote()
a2 = Counter.options(num_cpus=2, resources={"Custom2": 1}).remote()
a3 = Counter.options(num_cpus=3, resources={"Custom3": 1}).remote()
You can specify different resource requirements for tasks (but not for actor methods):
@@ -36,7 +36,7 @@ You can specify different resource requirements for tasks (but not for actor met
object_gpu_ids = g.remote()
assert ray.get(object_gpu_ids) == [0]
dynamic_object_gpu_ids = g._remote(args=[], num_cpus=1, num_gpus=1)
dynamic_object_gpu_ids = g.options(num_cpus=1, num_gpus=1).remote()
assert ray.get(dynamic_object_gpu_ids) == [0]
And vary the number of return values for tasks (and actor methods too):
@@ -47,10 +47,11 @@ And vary the number of return values for tasks (and actor methods too):
def f(n):
return list(range(n))
id1, id2 = f._remote(args=[2], num_return_vals=2)
id1, id2 = f.options(num_return_vals=2).remote(2)
assert ray.get(id1) == 0
assert ray.get(id2) == 1
Dynamic Custom Resources
------------------------
@@ -250,7 +251,7 @@ For example, you can instantiate and register a persistent actor as follows:
.. code-block:: python
counter = Counter._remote(name="CounterActor", detached=True)
counter = Counter.options(name="CounterActor", detached=True).remote()
The CounterActor will be kept alive even after the driver running above script
exits. Therefore it is possible to run the following script in a different
@@ -266,7 +267,7 @@ up after driver exits:
.. code-block:: python
Counter._remote(name="CounterActor")
Counter.options(name="CounterActor").remote()
However, creating a detached actor without name is not allowed because there
will be no way to retrieve the actor handle and the resource is leaked.
@@ -274,4 +275,4 @@ will be no way to retrieve the actor handle and the resource is leaked.
.. code-block:: python
# Can't do this!
Counter._remote(detached=True)
Counter.options(detached=True).remote()