Reducing test flakiness (#601)

Author: mvsusp

tests/integ/conftest.py

@@ -11,9 +11,25 @@
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
 from __future__ import absolute_import
+
+import os
+
 import pytest
 
 
+def create_sagemaker_local_network():
+    """
+    Docker has a known race condition which allows two parallel processes
+    to create a duplicated networks with the same name. This function
+    creates the network sagemaker-local beforehand, avoiding this issue
+    in CI.
+    """
+    os.system('docker network create sagemaker-local')
+
+
+create_sagemaker_local_network()
+
+
 @pytest.fixture(scope='session', params=['local', 'ml.c4.xlarge'])
 def instance_type(request):
     return request.param

tests/integ/test_tf.py

@@ -38,6 +38,7 @@ def tf_training_job(sagemaker_session, tf_full_version):
                                framework_version=tf_full_version,
                                training_steps=1,
                                evaluation_steps=1,
+                               checkpoint_path='/opt/ml/model',
                                hyperparameters={'input_tensor_name': 'inputs'},
                                train_instance_count=1,
                                train_instance_type='ml.c4.xlarge',
@@ -114,6 +115,7 @@ def test_tf_async(sagemaker_session):
                                role='SageMakerRole',
                                training_steps=1,
                                evaluation_steps=1,
+                               checkpoint_path='/opt/ml/model',
                                hyperparameters={'input_tensor_name': 'inputs'},
                                train_instance_count=1,
                                train_instance_type='ml.c4.xlarge',