Merge branch 'master' into keras_fn

Author: yangaws

CHANGELOG.rst

@@ -2,6 +2,11 @@
 CHANGELOG
 =========
 
+1.9.1dev
+========
+
+* bug-fix: Estimators: Fix serialization of single records
+
 1.9.0
 =====
 

src/sagemaker/amazon/common.py

@@ -29,7 +29,7 @@ def __init__(self, content_type='application/x-recordio-protobuf'):
 
     def __call__(self, array):
         if len(array.shape) == 1:
-            array.reshape(1, array.shape[0])
+            array = array.reshape(1, array.shape[0])
         assert len(array.shape) == 2, "Expecting a 1 or 2 dimensional array"
         buf = io.BytesIO()
         write_numpy_to_dense_tensor(buf, array)

tests/data/dummy_tensor

@@ -13,19 +13,15 @@
 from __future__ import absolute_import
 
 import gzip
-import io
 import json
 import os
 import pickle
 import sys
 
-import boto3
-import numpy as np
 import pytest
 
 import sagemaker
 from sagemaker.amazon.amazon_estimator import registry
-from sagemaker.amazon.common import write_numpy_to_dense_tensor
 from sagemaker.estimator import Estimator
 from sagemaker.utils import name_from_base
 from tests.integ import DATA_DIR
@@ -57,6 +53,7 @@ def test_byo_estimator(sagemaker_session, region):
 
     """
     image_name = registry(region) + "/factorization-machines:1"
+    training_data_path = os.path.join(DATA_DIR, 'dummy_tensor')
 
     with timeout(minutes=15):
         data_path = os.path.join(DATA_DIR, 'one_p_mnist', 'mnist.pkl.gz')
@@ -65,19 +62,11 @@ def test_byo_estimator(sagemaker_session, region):
         with gzip.open(data_path, 'rb') as f:
             train_set, _, _ = pickle.load(f, **pickle_args)
 
-        # take 100 examples for faster execution
-        vectors = np.array([t.tolist() for t in train_set[0][:100]]).astype('float32')
-        labels = np.where(np.array([t.tolist() for t in train_set[1][:100]]) == 0, 1.0, 0.0).astype('float32')
-
-        buf = io.BytesIO()
-        write_numpy_to_dense_tensor(buf, vectors, labels)
-        buf.seek(0)
-
-        bucket = sagemaker_session.default_bucket()
         prefix = 'test_byo_estimator'
         key = 'recordio-pb-data'
-        boto3.resource('s3').Bucket(bucket).Object(os.path.join(prefix, 'train', key)).upload_fileobj(buf)
-        s3_train_data = 's3://{}/{}/train/{}'.format(bucket, prefix, key)
+
+        s3_train_data = sagemaker_session.upload_data(path=training_data_path,
+                                                      key_prefix=os.path.join(prefix, 'train', key))
 
         estimator = Estimator(image_name=image_name,
                               role='SageMakerRole', train_instance_count=1,
@@ -111,6 +100,7 @@ def test_byo_estimator(sagemaker_session, region):
 def test_async_byo_estimator(sagemaker_session, region):
     image_name = registry(region) + "/factorization-machines:1"
     endpoint_name = name_from_base('byo')
+    training_data_path = os.path.join(DATA_DIR, 'dummy_tensor')
     training_job_name = ""
 
     with timeout(minutes=5):
@@ -120,19 +110,11 @@ def test_async_byo_estimator(sagemaker_session, region):
         with gzip.open(data_path, 'rb') as f:
             train_set, _, _ = pickle.load(f, **pickle_args)
 
-        # take 100 examples for faster execution
-        vectors = np.array([t.tolist() for t in train_set[0][:100]]).astype('float32')
-        labels = np.where(np.array([t.tolist() for t in train_set[1][:100]]) == 0, 1.0, 0.0).astype('float32')
-
-        buf = io.BytesIO()
-        write_numpy_to_dense_tensor(buf, vectors, labels)
-        buf.seek(0)
-
-        bucket = sagemaker_session.default_bucket()
         prefix = 'test_byo_estimator'
         key = 'recordio-pb-data'
-        boto3.resource('s3').Bucket(bucket).Object(os.path.join(prefix, 'train', key)).upload_fileobj(buf)
-        s3_train_data = 's3://{}/{}/train/{}'.format(bucket, prefix, key)
+
+        s3_train_data = sagemaker_session.upload_data(path=training_data_path,
+                                                      key_prefix=os.path.join(prefix, 'train', key))
 
         estimator = Estimator(image_name=image_name,
                               role='SageMakerRole', train_instance_count=1,

tests/integ/test_byo_estimator.py

@@ -66,15 +66,15 @@ def test_training_with_additional_hyperparameters(sagemaker_session, chainer_ful
 def test_attach_deploy(chainer_training_job, sagemaker_session):
     endpoint_name = 'test-chainer-attach-deploy-{}'.format(sagemaker_timestamp())
 
-    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, minutes=20):
+    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         estimator = Chainer.attach(chainer_training_job, sagemaker_session=sagemaker_session)
         predictor = estimator.deploy(1, 'ml.m4.xlarge', endpoint_name=endpoint_name)
         _predict_and_assert(predictor)
 
 
 def test_deploy_model(chainer_training_job, sagemaker_session):
     endpoint_name = 'test-chainer-deploy-model-{}'.format(sagemaker_timestamp())
-    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, minutes=20):
+    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         desc = sagemaker_session.sagemaker_client.describe_training_job(TrainingJobName=chainer_training_job)
         model_data = desc['ModelArtifacts']['S3ModelArtifacts']
         script_path = os.path.join(DATA_DIR, 'chainer_mnist', 'mnist.py')
@@ -93,7 +93,7 @@ def test_async_fit(sagemaker_session):
         print("Waiting to re-attach to the training job: %s" % training_job_name)
         time.sleep(20)
 
-    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, minutes=35):
+    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         print("Re-attaching now to: %s" % training_job_name)
         estimator = Chainer.attach(training_job_name=training_job_name, sagemaker_session=sagemaker_session)
         predictor = estimator.deploy(1, "ml.c4.xlarge", endpoint_name=endpoint_name)

tests/integ/test_chainer_train.py

@@ -41,7 +41,7 @@ def fixture_training_job(sagemaker_session, pytorch_full_version):
 def test_sync_fit_deploy(pytorch_training_job, sagemaker_session):
     # TODO: add tests against local mode when it's ready to be used
     endpoint_name = 'test-pytorch-sync-fit-attach-deploy{}'.format(sagemaker_timestamp())
-    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, minutes=20):
+    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         estimator = PyTorch.attach(pytorch_training_job, sagemaker_session=sagemaker_session)
         predictor = estimator.deploy(1, 'ml.c4.xlarge', endpoint_name=endpoint_name)
         data = numpy.zeros(shape=(1, 1, 28, 28), dtype=numpy.float32)

tests/integ/test_pytorch_train.py

@@ -34,7 +34,7 @@ def test_randomcutforest(sagemaker_session):
         rcf.fit(rcf.record_set(train_input))
 
     endpoint_name = name_from_base('randomcutforest')
-    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, minutes=20):
+    with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         model = RandomCutForestModel(rcf.model_data, role='SageMakerRole', sagemaker_session=sagemaker_session)
         predictor = model.deploy(1, 'ml.c4.xlarge', endpoint_name=endpoint_name)
 

tests/integ/test_randomcutforest.py

@@ -13,20 +13,18 @@
 from __future__ import absolute_import
 
 import gzip
-import io
 import json
 import os
 import pickle
 import sys
 import time
 
-import boto3
 import numpy as np
 import pytest
 
 from sagemaker import KMeans, LDA, RandomCutForest
 from sagemaker.amazon.amazon_estimator import registry
-from sagemaker.amazon.common import read_records, write_numpy_to_dense_tensor
+from sagemaker.amazon.common import read_records
 from sagemaker.chainer import Chainer
 from sagemaker.estimator import Estimator
 from sagemaker.mxnet.estimator import MXNet
@@ -344,7 +342,7 @@ def test_attach_tuning_pytorch(sagemaker_session):
 
     attached_tuner = HyperparameterTuner.attach(tuning_job_name, sagemaker_session=sagemaker_session)
     best_training_job = tuner.best_training_job()
-    with timeout_and_delete_endpoint_by_name(best_training_job, sagemaker_session, minutes=20):
+    with timeout_and_delete_endpoint_by_name(best_training_job, sagemaker_session):
         predictor = attached_tuner.deploy(1, 'ml.c4.xlarge')
         data = np.zeros(shape=(1, 1, 28, 28), dtype=np.float32)
         predictor.predict(data)
@@ -368,6 +366,7 @@ def test_tuning_byo_estimator(sagemaker_session):
     Default predictor is updated with json serializer and deserializer.
     """
     image_name = registry(sagemaker_session.boto_session.region_name) + '/factorization-machines:1'
+    training_data_path = os.path.join(DATA_DIR, 'dummy_tensor')
 
     with timeout(minutes=15):
         data_path = os.path.join(DATA_DIR, 'one_p_mnist', 'mnist.pkl.gz')
@@ -376,19 +375,10 @@ def test_tuning_byo_estimator(sagemaker_session):
         with gzip.open(data_path, 'rb') as f:
             train_set, _, _ = pickle.load(f, **pickle_args)
 
-        # take 100 examples for faster execution
-        vectors = np.array([t.tolist() for t in train_set[0][:100]]).astype('float32')
-        labels = np.where(np.array([t.tolist() for t in train_set[1][:100]]) == 0, 1.0, 0.0).astype('float32')
-
-        buf = io.BytesIO()
-        write_numpy_to_dense_tensor(buf, vectors, labels)
-        buf.seek(0)
-
-        bucket = sagemaker_session.default_bucket()
         prefix = 'test_byo_estimator'
         key = 'recordio-pb-data'
-        boto3.resource('s3').Bucket(bucket).Object(os.path.join(prefix, 'train', key)).upload_fileobj(buf)
-        s3_train_data = 's3://{}/{}/train/{}'.format(bucket, prefix, key)
+        s3_train_data = sagemaker_session.upload_data(path=training_data_path,
+                                                      key_prefix=os.path.join(prefix, 'train', key))
 
         estimator = Estimator(image_name=image_name,
                               role='SageMakerRole', train_instance_count=1,

tests/integ/test_tuner.py

@@ -60,7 +60,7 @@ def handler(signum, frame):
 
 
 @contextmanager
-def timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, seconds=0, minutes=35, hours=0):
+def timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session, seconds=0, minutes=45, hours=0):
     with timeout(seconds=seconds, minutes=minutes, hours=hours) as t:
         no_errors = False
         try:

tests/integ/timeout.py

@@ -32,6 +32,16 @@ def test_serializer():
         assert record.features["values"].float64_tensor.values == expected
 
 
+def test_serializer_accepts_one_dimensional_array():
+    s = numpy_to_record_serializer()
+    array_data = [1.0, 2.0, 3.0]
+    buf = s(np.array(array_data))
+    record_data = next(_read_recordio(buf))
+    record = Record()
+    record.ParseFromString(record_data)
+    assert record.features["values"].float64_tensor.values == array_data
+
+
 def test_deserializer():
     array_data = [[1.0, 2.0, 3.0], [10.0, 20.0, 30.0]]
     s = numpy_to_record_serializer()