Enable standalone executor for single-GPU training #45913
Conversation
|
你的PR提交成功，感谢你对开源项目的贡献! |
| @@ -1555,23 +1555,32 @@ def _can_use_interpreter_core(program, place): | |||
| place, core.CustomPlace): | |||
| return False | |||
|
|
|||
| use_standalone_executor_for_compiled_program = os.environ.get( | |||
| use_standalone_executor_for_distribution = os.environ.get( | |||
There was a problem hiding this comment.
NIT, why change to for distribution?
There was a problem hiding this comment.
After this PR, the FLAGS_CONVERT_GRAPH_TO_PROGRAM is no longer control whether to run Compiled Program with Standalone Executor, but control whether to run distribution.
It is better to change the FLAGS name at the same time. However, considering that it is just a temporary FLAGS and QAs have been set it in their testing environment, keeping the same FLAGS name can avoid the requirement for QA to reset FLAGS when they try to test the distributed scenario of Standalone Executor.
PR types
New features
PR changes
Others
Describe
PR #41179 has enabled Standalone Executor for Program running, this PR enabled it for Compiled Program. After this PR, all single-GPU training using old Executor or Parallel Executor (PE) are set to use Standalone Executor by default.
The supporting for distributed running is ongoing.