def tfs_predictor_with_model_and_entry_point_and_dependencies(
instance_type, sagemaker_session, tf_full_version):
endpoint_name = sagemaker.utils.unique_name_from_base(
'sagemaker-tensorflow-serving')
model_data = sagemaker_session.upload_data(
path=os.path.join(tests.integ.DATA_DIR,
'tensorflow-serving-test-model.tar.gz'),
key_prefix='tensorflow-serving/models')
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(
endpoint_name, sagemaker_session):
entry_point = os.path.join(
tests.integ.DATA_DIR,
'tfs/tfs-test-entrypoint-and-dependencies/inference.py')
dependencies = [
os.path.join(
tests.integ.DATA_DIR,
'tfs/tfs-test-entrypoint-and-dependencies/dependency.py')
]
model = Model(entry_point=entry_point,
model_data=model_data,
role='SageMakerRole',
dependencies=dependencies,
framework_version=tf_full_version,
sagemaker_session=sagemaker_session)
predictor = model.deploy(1, instance_type, endpoint_name=endpoint_name)
yield predictor
def tfs_predictor_with_model_and_entry_point_and_dependencies(
sagemaker_local_session, tf_full_version):
endpoint_name = sagemaker.utils.unique_name_from_base(
"sagemaker-tensorflow-serving")
entry_point = os.path.join(
tests.integ.DATA_DIR,
"tfs/tfs-test-entrypoint-and-dependencies/inference.py")
dependencies = [
os.path.join(tests.integ.DATA_DIR,
"tfs/tfs-test-entrypoint-and-dependencies/dependency.py")
]
model_data = "file://" + os.path.join(
tests.integ.DATA_DIR, "tensorflow-serving-test-model.tar.gz")
model = Model(
entry_point=entry_point,
model_data=model_data,
role="SageMakerRole",
dependencies=dependencies,
framework_version=tf_full_version,
sagemaker_session=sagemaker_local_session,
)
predictor = model.deploy(1, "local", endpoint_name=endpoint_name)
try:
yield predictor
finally:
predictor.delete_endpoint()
def test_tfs_model_image_accelerator_not_supported(sagemaker_session):
model = Model(
"s3://some/data.tar.gz",
role=ROLE,
framework_version="1.13.1",
sagemaker_session=sagemaker_session,
)
# assert error is not raised
model.deploy(instance_type="ml.c4.xlarge",
initial_instance_count=1,
accelerator_type="ml.eia1.medium")
model = Model(
"s3://some/data.tar.gz",
role=ROLE,
framework_version="1.14",
sagemaker_session=sagemaker_session,
)
# assert error is not raised
model.deploy(instance_type="ml.c4.xlarge", initial_instance_count=1)
with pytest.raises(AttributeError) as e:
model.deploy(
instance_type="ml.c4.xlarge",
accelerator_type="ml.eia1.medium",
initial_instance_count=1,
)
assert str(e.value) == "The TensorFlow version 1.14 doesn't support EIA."
def test_tfs_model_with_dependencies(repack_model, model_code_key_prefix,
sagemaker_session, tf_version):
model = Model(
"s3://some/data.tar.gz",
entry_point="train.py",
dependencies=["src", "lib"],
role=ROLE,
framework_version=tf_version,
image="my-image",
sagemaker_session=sagemaker_session,
)
model.prepare_container_def(INSTANCE_TYPE)
model_code_key_prefix.assert_called_with(model.key_prefix, model.name,
model.image)
repack_model.assert_called_with(
"train.py",
None,
["src", "lib"],
"s3://some/data.tar.gz",
"s3://my_bucket/key-prefix/model.tar.gz",
sagemaker_session,
)
def test_tfs_model_with_custom_image(sagemaker_session, tf_version):
model = Model("s3://some/data.tar.gz",
role=ROLE,
framework_version=tf_version,
image='my-image',
sagemaker_session=sagemaker_session)
cdef = model.prepare_container_def(INSTANCE_TYPE)
assert cdef['Image'] == 'my-image'
def test_model_prepare_container_def_no_instance_type_or_image():
model = Model("s3://some/data.tar.gz", role=ROLE)
with pytest.raises(ValueError) as e:
model.prepare_container_def()
expected_msg = "Must supply either an instance type (for choosing CPU vs GPU) or an image URI."
assert expected_msg in str(e)
def test_tfs_model_with_log_level(sagemaker_session, tf_version):
model = Model("s3://some/data.tar.gz",
role=ROLE,
framework_version=tf_version,
container_log_level=logging.INFO,
sagemaker_session=sagemaker_session)
cdef = model.prepare_container_def(INSTANCE_TYPE)
assert cdef['Environment'] == {Model.LOG_LEVEL_PARAM_NAME: 'info'}
def test_tfs_model(sagemaker_session, tf_version):
model = Model("s3://some/data.tar.gz", role=ROLE, framework_version=tf_version,
sagemaker_session=sagemaker_session)
cdef = model.prepare_container_def(INSTANCE_TYPE)
assert cdef['Image'].endswith('sagemaker-tensorflow-serving:{}-cpu'.format(tf_version))
assert cdef['Environment'] == {}
predictor = model.deploy(INSTANCE_COUNT, INSTANCE_TYPE)
assert isinstance(predictor, Predictor)
def main():
parser = argparse.ArgumentParser()
parser.add_argument('model_data', help='s3 path to model.tar.gz')
parser.add_argument('--instance_type', default='ml.m5.large')
parser.add_argument('--role', default=os.environ['SAGEMAKER_ROLE'])
args = parser.parse_args()
model = Model(args.model_data, args.role, framework_version='1.12.0')
model.deploy(initial_instance_count=1, instance_type=args.instance_type)
def test_enabling_data_capture_on_endpoint_shows_correct_data_capture_status(
sagemaker_session, tf_full_version
):
endpoint_name = unique_name_from_base("sagemaker-tensorflow-serving")
model_data = sagemaker_session.upload_data(
path=os.path.join(tests.integ.DATA_DIR, "tensorflow-serving-test-model.tar.gz"),
key_prefix="tensorflow-serving/models",
)
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
model = Model(
model_data=model_data,
role=ROLE,
framework_version=tf_full_version,
sagemaker_session=sagemaker_session,
)
predictor = model.deploy(
initial_instance_count=INSTANCE_COUNT,
instance_type=INSTANCE_TYPE,
endpoint_name=endpoint_name,
)
endpoint_desc = sagemaker_session.sagemaker_client.describe_endpoint(
EndpointName=predictor.endpoint
)
endpoint_config_desc = sagemaker_session.sagemaker_client.describe_endpoint_config(
EndpointConfigName=endpoint_desc["EndpointConfigName"]
)
assert endpoint_config_desc.get("DataCaptureConfig") is None
predictor.enable_data_capture()
# Wait for endpoint to finish updating
# Endpoint update takes ~7min. 40 retries * 30s sleeps = 20min timeout
for _ in retries(
max_retry_count=40,
exception_message_prefix="Waiting for 'InService' endpoint status",
seconds_to_sleep=30,
):
new_endpoint = sagemaker_session.sagemaker_client.describe_endpoint(
EndpointName=predictor.endpoint
)
if new_endpoint["EndpointStatus"] == "InService":
break
endpoint_desc = sagemaker_session.sagemaker_client.describe_endpoint(
EndpointName=predictor.endpoint
)
endpoint_config_desc = sagemaker_session.sagemaker_client.describe_endpoint_config(
EndpointConfigName=endpoint_desc["EndpointConfigName"]
)
assert endpoint_config_desc["DataCaptureConfig"]["EnableCapture"]
def test_tfs_model_image_accelerator(sagemaker_session, tf_version):
model = Model(
"s3://some/data.tar.gz",
role=ROLE,
framework_version=tf_version,
sagemaker_session=sagemaker_session,
)
cdef = model.prepare_container_def(INSTANCE_TYPE, accelerator_type=ACCELERATOR_TYPE)
assert cdef["Image"].endswith("sagemaker-tensorflow-serving-eia:{}-cpu".format(tf_version))
predictor = model.deploy(INSTANCE_COUNT, INSTANCE_TYPE)
assert isinstance(predictor, Predictor)
def deploy(model_data, endpoint_name):
"""
Deploys a SageMaker endpoint for a trained model.
:param model_data: S3 location of the model artifacts.
:param endpoint_name: Name to assign to the SageMaker endpoint.
"""
# Create the model endpoint.
model = Model(model_data=model_data,
role=os.getenv("SAGEMAKER_ROLE"),
image=os.getenv("SAGEMAKER_IMAGE_URI"))
model.deploy(1, 'ml.t2.medium', endpoint_name=endpoint_name)
def tfs_predictor(instance_type, sagemaker_session, tf_full_version):
endpoint_name = sagemaker.utils.name_from_base(
'sagemaker-tensorflow-serving')
model_data = sagemaker_session.upload_data(
path='tests/data/tensorflow-serving-test-model.tar.gz',
key_prefix='tensorflow-serving/models')
with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
model = Model(model_data=model_data,
role='SageMakerRole',
framework_version=tf_full_version,
sagemaker_session=sagemaker_session)
predictor = model.deploy(1, instance_type, endpoint_name=endpoint_name)
yield predictor
def tfs_predictor(sagemaker_session, tf_full_version):
endpoint_name = sagemaker.utils.unique_name_from_base("sagemaker-tensorflow-serving")
model_data = sagemaker_session.upload_data(
path=os.path.join(tests.integ.DATA_DIR, "tensorflow-serving-test-model.tar.gz"),
key_prefix="tensorflow-serving/models",
)
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
model = Model(
model_data=model_data,
role="SageMakerRole",
framework_version=tf_full_version,
sagemaker_session=sagemaker_session,
)
predictor = model.deploy(1, "ml.c5.xlarge", endpoint_name=endpoint_name)
yield predictor
def _create_tfs_model(self,
role=None,
vpc_config_override=VPC_CONFIG_DEFAULT,
entry_point=None,
source_dir=None,
dependencies=None,
**kwargs):
"""Placeholder docstring"""
# remove image kwarg
if "image" in kwargs:
image = kwargs["image"]
kwargs = {k: v for k, v in kwargs.items() if k != "image"}
else:
image = None
return Model(model_data=self.model_data,
role=role,
image=(image or self.image_name),
name=self._current_job_name,
container_log_level=self.container_log_level,
framework_version=utils.get_short_version(
self.framework_version),
sagemaker_session=self.sagemaker_session,
vpc_config=self.get_vpc_config(vpc_config_override),
entry_point=entry_point,
source_dir=source_dir,
dependencies=dependencies,
enable_network_isolation=self.enable_network_isolation(),
**kwargs)
def tfs_predictor_with_accelerator(sagemaker_session, ei_tf_full_version, cpu_instance_type):
endpoint_name = sagemaker.utils.unique_name_from_base("sagemaker-tensorflow-serving")
model_data = sagemaker_session.upload_data(
path=os.path.join(tests.integ.DATA_DIR, "tensorflow-serving-test-model.tar.gz"),
key_prefix="tensorflow-serving/models",
)
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
model = Model(
model_data=model_data,
role="SageMakerRole",
framework_version=ei_tf_full_version,
sagemaker_session=sagemaker_session,
)
predictor = model.deploy(
1, cpu_instance_type, endpoint_name=endpoint_name, accelerator_type="ml.eia1.medium"
)
yield predictor
def test_tfs_model_with_entry_point(repack_model, model_code_key_prefix,
sagemaker_session, tf_version):
model = Model("s3://some/data.tar.gz",
entry_point='train.py',
role=ROLE,
framework_version=tf_version,
image='my-image',
sagemaker_session=sagemaker_session)
model.prepare_container_def(INSTANCE_TYPE)
model_code_key_prefix.assert_called_with(model.key_prefix, model.name,
model.image)
repack_model.assert_called_with('train.py', None, [],
's3://some/data.tar.gz',
's3://my_bucket/key-prefix/model.tar.gz',
sagemaker_session)
def _create_tfs_model(self, role=None, vpc_config_override=VPC_CONFIG_DEFAULT):
return Model(model_data=self.model_data,
role=role,
image=self.image_name,
name=self._current_job_name,
container_log_level=self.container_log_level,
framework_version=self.framework_version,
sagemaker_session=self.sagemaker_session,
vpc_config=self.get_vpc_config(vpc_config_override))
def tfs_predictor_with_accelerator(sagemaker_session, tf_full_version):
endpoint_name = sagemaker.utils.unique_name_from_base(
"sagemaker-tensorflow-serving")
instance_type = 'ml.c4.large'
accelerator_type = 'ml.eia1.medium'
model_data = sagemaker_session.upload_data(
path='tests/data/tensorflow-serving-test-model.tar.gz',
key_prefix='tensorflow-serving/models')
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(
endpoint_name, sagemaker_session):
model = Model(model_data=model_data,
role='SageMakerRole',
framework_version=tf_full_version,
sagemaker_session=sagemaker_session)
predictor = model.deploy(1,
instance_type,
endpoint_name=endpoint_name,
accelerator_type=accelerator_type)
yield predictor
def tfs_predictor_with_model_and_entry_point_same_tar(instance_type,
sagemaker_session,
tf_full_version, tmpdir):
endpoint_name = sagemaker.utils.unique_name_from_base(
'sagemaker-tensorflow-serving')
model_tar = tar_dir(
os.path.join(tests.integ.DATA_DIR,
'tfs/tfs-test-model-with-inference'), tmpdir)
model_data = sagemaker_session.upload_data(
path=model_tar, key_prefix='tensorflow-serving/models')
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(
endpoint_name, sagemaker_session):
model = Model(model_data=model_data,
role='SageMakerRole',
framework_version=tf_full_version,
sagemaker_session=sagemaker_session)
predictor = model.deploy(1, instance_type, endpoint_name=endpoint_name)
yield predictor
def tfs_predictor_with_model_and_entry_point_same_tar(sagemaker_local_session,
tf_full_version, tmpdir):
endpoint_name = sagemaker.utils.unique_name_from_base(
"sagemaker-tensorflow-serving")
model_tar = tar_dir(
os.path.join(tests.integ.DATA_DIR,
"tfs/tfs-test-model-with-inference"), tmpdir)
model = Model(
model_data="file://" + model_tar,
role="SageMakerRole",
framework_version=tf_full_version,
sagemaker_session=sagemaker_local_session,
)
predictor = model.deploy(1, "local", endpoint_name=endpoint_name)
try:
yield predictor
finally:
predictor.delete_endpoint()
def Deploy_Model(payload, update=False):
''' SERVICE NUMBER 2. The function deploys the model as a endpoint
payload = {model_data:model_data}
sample model_data : 's3://fypcementbucket/models/model_2021_2_19/sagemaker-tensorflow-scriptmode-2021-02-21-13-37-05-805/output/model.tar.gz'
'''
from sagemaker.tensorflow.serving import Model
import boto3
client = boto3.client('sagemaker')
#Try deleting the previous configurations if present
try:
client.delete_endpoint_config(
EndpointConfigName=
'sagemaker-tensorflow-serving-2021-02-21-fypmodel-endpoint')
except:
pass
model = Model(
model_data=
's3://fypcementbucket/models/model_3_16_14_52_30/sagemaker-tensorflow-scriptmode-2021-03-16-14-52-31-100/output/model.tar.gz',
role=
"arn:aws:iam::968710761052:role/service-role/AmazonSageMaker-ExecutionRole-20210205T194406",
framework_version='1.12.0')
newThread = threading.Thread(target=Thread_Handler)
newThread.start()
predictor = model.deploy(
initial_instance_count=1,
instance_type='ml.c5.xlarge',
endpoint_name=
'sagemaker-tensorflow-serving-2021-02-21-fypmodel-endpoint',
update_endpoint=update)
Get_ModelStatus()
def Deploy_Model(payload, update=False):
''' SERVICE NUMBER 2. The function deploys the model as a endpoint
payload = {model_data:model_data}
sample model_data : 's3://fypcementbucket/models/model_2021_2_19/sagemaker-tensorflow-scriptmode-2021-02-21-13-37-05-805/output/model.tar.gz'
'''
from sagemaker.tensorflow.serving import Model
model = Model(
model_data=payload['model_data'],
role="arn:aws:iam::968710761052:role/service-role/AmazonSageMaker-ExecutionRole-20210205T194406",
framework_version='1.12.0'
)
newThread = threading.Thread(target=Thread_Handler)
newThread.start()
predictor = model.deploy(initial_instance_count=1, instance_type='ml.c5.xlarge',
endpoint_name='sagemaker-tensorflow-serving-2021-02-21-fypmodel-endpoint', update_endpoint=update)
Get_ModelStatus()
def _create_tfs_model(self,
role=None,
vpc_config_override=VPC_CONFIG_DEFAULT,
entry_point=None,
source_dir=None,
dependencies=None,
**kwargs):
"""Placeholder docstring"""
return Model(model_data=self.model_data,
role=role,
container_log_level=self.container_log_level,
framework_version=utils.get_short_version(
self.framework_version),
sagemaker_session=self.sagemaker_session,
vpc_config=self.get_vpc_config(vpc_config_override),
entry_point=entry_point,
source_dir=source_dir,
dependencies=dependencies,
**kwargs)
sage_session = sagemaker.session.Session()
s3_bucket = "bonhomme-snake"
role = sagemaker.get_execution_role()
s3_client = boto3.client('s3')
s3_client.upload_file("RLlibEnv/inference/model.tar.gz", s3_bucket,
"battlesnake-aws/pretrainedmodels/model.tar.gz")
model_data = "s3://{}/battlesnake-aws/pretrainedmodels/model.tar.gz".format(
s3_bucket)
print("Make an endpoint with {}".format(model_data))
model = Model(
model_data=model_data,
role=role,
entry_point="inference.py",
source_dir='RLlibEnv/inference/inference_src',
framework_version='2.1.0',
name="battlesnake-rllib",
)
# Deploy an inference endpoint
predictor = model.deploy(initial_instance_count=1,
instance_type=endpoint_instance_type,
endpoint_name='battlesnake-endpoint')
state = np.zeros(shape=(1, 21, 21, 6), dtype=np.float32).tolist()
health_dict = {0: 50, 1: 50}
json = {
"turn": 4,
"board": {
from sagemaker.tensorflow.serving import Model
model = Model(model_data='s3://<AWS_BUCKET>/<S3_MODEL_PATH>',
framework_version='1.13',
role='<AWS_ROLE>')
predictor = model.deploy(initial_instance_count=1,
instance_type='ml.c5.xlarge')
#nbRatingsTest = prepdata_result["Parameters"]["NbRatingsTest"]
#testData = prepdata_result["Parameters"]["TestData"]
with open(cf_configuration_filepath) as cf_configuration_file:
cf_configuration = json.load(cf_configuration_file)
commit_id = cf_configuration["Parameters"]["CommitID"]
timestamp = cf_configuration["Parameters"]["Timestamp"]
model_data = cf_configuration["Parameters"]["ModelData"]
stage = cf_configuration["Parameters"]["Environment"]
sagemaker_session = sagemaker.Session()
role = get_execution_role()
tensorflow_serving_model = Model(model_data=model_data,
role=role,
framework_version='1.13',
sagemaker_session=sagemaker_session)
predictor = tensorflow_serving_model.deploy(initial_instance_count=1,
instance_type='ml.t2.medium')
result = predictor.predict(json_str)
result = result['predictions']
pred = np.array(result)
pred = pred.argmax(axis=1).tolist()
y_test_pred = y_test.argmax(axis=1).tolist()
if ((stage == 'qa') or (stage == 'prod')):
predictor.delete_endpoint()
match = 0
def test_disabling_data_capture_on_endpoint_shows_correct_data_capture_status(
sagemaker_session, tf_full_version):
endpoint_name = unique_name_from_base("sagemaker-tensorflow-serving")
model_data = sagemaker_session.upload_data(
path=os.path.join(tests.integ.DATA_DIR,
"tensorflow-serving-test-model.tar.gz"),
key_prefix="tensorflow-serving/models",
)
with tests.integ.timeout.timeout_and_delete_endpoint_by_name(
endpoint_name, sagemaker_session):
model = Model(
model_data=model_data,
role=ROLE,
framework_version=tf_full_version,
sagemaker_session=sagemaker_session,
)
destination_s3_uri = os.path.join("s3://",
sagemaker_session.default_bucket(),
endpoint_name, "custom")
predictor = model.deploy(
initial_instance_count=INSTANCE_COUNT,
instance_type=INSTANCE_TYPE,
endpoint_name=endpoint_name,
data_capture_config=DataCaptureConfig(
enable_capture=True,
sampling_percentage=CUSTOM_SAMPLING_PERCENTAGE,
destination_s3_uri=destination_s3_uri,
capture_options=CUSTOM_CAPTURE_OPTIONS,
csv_content_types=CUSTOM_CSV_CONTENT_TYPES,
json_content_types=CUSTOM_JSON_CONTENT_TYPES,
sagemaker_session=sagemaker_session,
),
)
endpoint_desc = sagemaker_session.sagemaker_client.describe_endpoint(
EndpointName=predictor.endpoint)
endpoint_config_desc = sagemaker_session.sagemaker_client.describe_endpoint_config(
EndpointConfigName=endpoint_desc["EndpointConfigName"])
assert endpoint_config_desc["DataCaptureConfig"]["EnableCapture"]
assert (endpoint_config_desc["DataCaptureConfig"]
["InitialSamplingPercentage"] == CUSTOM_SAMPLING_PERCENTAGE)
assert endpoint_config_desc["DataCaptureConfig"]["CaptureOptions"] == [
{
"CaptureMode": "Input"
}
]
assert (endpoint_config_desc["DataCaptureConfig"]
["CaptureContentTypeHeader"]["CsvContentTypes"] ==
CUSTOM_CSV_CONTENT_TYPES)
assert (endpoint_config_desc["DataCaptureConfig"]
["CaptureContentTypeHeader"]["JsonContentTypes"] ==
CUSTOM_JSON_CONTENT_TYPES)
predictor.disable_data_capture()
# Wait for endpoint to finish updating
# Endpoint update takes ~7min. 25 retries * 60s sleeps = 25min timeout
for _ in retries(
max_retry_count=25,
exception_message_prefix=
"Waiting for 'InService' endpoint status",
seconds_to_sleep=60,
):
new_endpoint = sagemaker_session.sagemaker_client.describe_endpoint(
EndpointName=predictor.endpoint)
if new_endpoint["EndpointStatus"] == "InService":
break
endpoint_desc = sagemaker_session.sagemaker_client.describe_endpoint(
EndpointName=predictor.endpoint)
endpoint_config_desc = sagemaker_session.sagemaker_client.describe_endpoint_config(
EndpointConfigName=endpoint_desc["EndpointConfigName"])
assert not endpoint_config_desc["DataCaptureConfig"]["EnableCapture"]
Exported된 SavedModel은 각 그래프에 correspond하는 tag-set가 있어야한다. SavedModel을 export하거나 save할때 specify해야한다. SavedModel의 tag-sets를 inspect하려면 SavedModel CLI를 사용. !!! https://www.tensorflow.org/programmers_guide/saved_model_cli !!! (404 에러) -> 리뷰: C C. SavedModel CLI https://www.tensorflow.org/guide/saved_model#details_of_the_savedmodel_command_line_interface #################### [Stack Overflow 답변] There are two APIs for deploying TensorFlow models: tensorflow.Model and tensorflow.serving.Model. It isn't clear from the code-snippet which one you're using, but the SageMaker docs recommend the latter deploying from pre-existing s3 artifacts: from sagemaker.tensorflow.serving import Model model = Model(model_data='s3://mybucket/model.tar.gz', role='MySageMakerRole') predictor = model.deploy(initial_instance_count=1, instance_type='ml.c5.xlarge') # Reference: https://github.com/aws/sagemaker-python-sdk/blob/c919e4dee3a00243f0b736af93fb156d17b04796/src/sagemaker/tensorflow/deploying_tensorflow_serving.rst#deploying-directly-from-model-artifacts If you haven't specified an image argument for tensorflow.Model, SageMaker should be using the default TensorFlow serving image (seems like "../tensorflow-inference"). image (str) – A Docker image URI (default: None). If not specified, a default image for TensorFlow Serving will be used. If all of this seems needlessly complex to you, I'm working on a platform that makes this set up a single line of code -- I'd love for you to try it, dm me at https://twitter.com/yoavz_.
def test_tfs_model_image_accelerator(sagemaker_session, tf_version):
model = Model("s3://some/data.tar.gz", role=ROLE, framework_version=tf_version,
sagemaker_session=sagemaker_session)
with pytest.raises(ValueError):
model.prepare_container_def(INSTANCE_TYPE, accelerator_type=ACCELERATOR_TYPE)