wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-09-09 11:32:43 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Ray SGD] use_local flag + Worker group abstraction (#10539)

Browse Source 
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
This commit is contained in:
Amog Kamsetty
2020-09-15 11:58:57 -07:00
committed by GitHub
co-authored by Richard Liaw
parent 0865d68466
commit d5a7c53908
15 changed files with 1025 additions and 426 deletions
Download Patch File Download Diff File Expand all files Collapse all files
doc/source/raysgd/raysgd_pytorch.rst
+3 -3
View File
@@ -32,7 +32,7 @@ The :ref:`ref-torch-trainer` can be constructed from a custom :ref:`ref-torch-o
:start-after: __torch_operator_start__
:end-before: __torch_operator_end__
Under the hood, ``TorchTrainer`` will create *replicas* of your model (controlled by ``num_workers``), each of which is managed by a Ray actor. One of the replicas will be on the main process, which can simplify the debugging and logging experience.
Under the hood, ``TorchTrainer`` will create *replicas* of your model (controlled by ``num_workers``), each of which is managed by a Ray actor.
Before instantiating the trainer, first start or connect to a Ray cluster:
@@ -288,8 +288,8 @@ However, if you have these creator functions already and do not want to change y
.. literalinclude:: ../../../python/ray/util/sgd/torch/examples/raysgd_torch_signatures.py
:language: python
:start-after: __backwards_compat__start
:end-before: __backwards_compat_end
:start-after: __backwards_compat_start__
:end-before: __backwards_compat_end__
Initialization Functions
------------------------