experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fix CollectiveOpTest.testCheckHealth failures in OSS

Browse Source 
In OSS fail_fast is false, so the error could be DeadlineExceededError.

PiperOrigin-RevId: 338743502
Change-Id: Ib49e8d6337882007ab5b91dde08129867b8b6aac
This commit is contained in:
Ran Chen 2020-10-23 14:18:43 -07:00 committed by TensorFlower Gardener
parent 7bf79598cb
commit ddfae909ed
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorflow/python/kernel_tests/collective_ops_multi_worker_test.py
View File

@ -80,7 +80,7 @@ class CollectiveOpTest(test.TestCase):
]:
context.context().check_collective_ops_peer_health(
task, timeout_in_ms=1000)
except errors.UnavailableError:
except (errors.UnavailableError, errors.DeadlineExceededError):
continue
break
multi_process_runner.get_barrier().wait()