keras_dnn_correctness_test_gpu doesn't seem flaky.

$ bazel test --config=cuda --runs_per_test=167 //third_party/tensorflow/python/keras/distribute:keras_dnn_correctness_test_gpu
//third_party/tensorflow/python/keras/distribute:keras_dnn_correctness_test_gpu PASSED in 840.3s
  Stats over 3173 runs: max = 840.3s, min = 64.5s, avg = 636.3s, dev = 139.8s

Executed 1 out of 1 test: 1 test passes.

PiperOrigin-RevId: 260229842
This commit is contained in:
Igor Saprykin 2019-07-26 15:54:54 -07:00 committed by TensorFlower Gardener
parent 8440153f82
commit f8c2d15c52

View File

@ -168,8 +168,6 @@ distribute_py_test(
"multi_and_single_gpu",
"no_rocm", # times out on ROCm
"no_windows_gpu",
# TODO(b/134764123): Re-enable this test.
"notap",
"notsan",
],
deps = [