change: enable Neo integ tests

src/sagemaker/model.py

@@ -304,8 +304,8 @@ def compile(
                 https://docs.aws.amazon.com/sagemaker/latest/dg/API_OutputConfig.html.
             input_shape (dict): Specifies the name and shape of the expected
                 inputs for your trained model in json dictionary form, for
-                example: {'data':[1,3,1024,1024]}, or {'var1': [1,1,28,28],
-                'var2':[1,1,28,28]}
+                example: {'data': [1,3,1024,1024]}, or {'var1': [1,1,28,28],
+                'var2': [1,1,28,28]}
             output_path (str): Specifies where to store the compiled model
             role (str): Execution role
             tags (list[dict]): List of tags for labeling a compilation job. For
@@ -436,7 +436,8 @@ def deploy(
 
         compiled_model_suffix = "-".join(instance_type.split(".")[:-1])
         if self._is_compiled_model:
-            self.name += compiled_model_suffix
+            name_prefix = self.name or utils.name_from_image(self.image)
+            self.name = "{}{}".format(name_prefix, compiled_model_suffix)
 
         self._create_sagemaker_model(instance_type, accelerator_type, tags)
         production_variant = sagemaker.production_variant(

tests/conftest.py

@@ -286,7 +286,7 @@ def alternative_cpu_instance_type(sagemaker_session, request):
 
 @pytest.fixture(scope="session")
 def cpu_instance_family(cpu_instance_type):
-    "_".join(cpu_instance_type.split(".")[0:2])
+    return "_".join(cpu_instance_type.split(".")[0:2])
 
 
 def pytest_generate_tests(metafunc):

tests/data/mxnet_mnist/mnist_neo.py

@@ -20,8 +20,6 @@
 import mxnet as mx
 import numpy as np
 
-from sagemaker_mxnet_container.training_utils import scheduler_host
-
 
 def load_data(path):
     with gzip.open(find_file(path, "labels.gz")) as flbl:
@@ -112,11 +110,10 @@ def neo_preprocess(payload, content_type):
     if content_type != "application/vnd+python.numpy+binary":
         raise RuntimeError("Content type must be application/vnd+python.numpy+binary")
 
-    f = io.BytesIO(payload)
-    return np.load(f)
+    return np.asarray(json.loads(payload.decode("utf-8")))
 
 
-### NOTE: this function cannot use MXNet
+# NOTE: this function cannot use MXNet
 def neo_postprocess(result):
     logging.info("Invoking user-defined post-processing function")
 
@@ -131,19 +128,10 @@ def neo_postprocess(result):
     return response_body, content_type
 
 
-def save(model_dir, model):
-    model.symbol.save(os.path.join(model_dir, "model-symbol.json"))
-    model.save_params(os.path.join(model_dir, "model-0000.params"))
-
-    signature = [
-        {"name": data_desc.name, "shape": [dim for dim in data_desc.shape]}
-        for data_desc in model.data_shapes
-    ]
-    with open(os.path.join(model_dir, "model-shapes.json"), "w") as f:
-        json.dump(signature, f)
-
-
 if __name__ == "__main__":
+    # Import here to prevent import during serving
+    from sagemaker_mxnet_container.training_utils import scheduler_host, save
+
     parser = argparse.ArgumentParser()
 
     parser.add_argument("--batch-size", type=int, default=100)

tests/integ/test_neo_mxnet.py

@@ -12,27 +12,30 @@
 # language governing permissions and limitations under the License.
 from __future__ import absolute_import
 
-import numpy
 import os
+
+import numpy
 import pytest
+
 from sagemaker.mxnet.estimator import MXNet
 from sagemaker.mxnet.model import MXNetModel
-from sagemaker.utils import sagemaker_timestamp
+from sagemaker.utils import unique_name_from_base
 from tests.integ import DATA_DIR, PYTHON_VERSION, TRAINING_DEFAULT_TIMEOUT_MINUTES
 from tests.integ.timeout import timeout, timeout_and_delete_endpoint_by_name
-import time
+
+NEO_MXNET_VERSION = "1.4.1"  # Neo doesn't support MXNet 1.6 yet.
 
 
 @pytest.fixture(scope="module")
-def mxnet_training_job(sagemaker_session, mxnet_full_version, cpu_instance_type):
+def mxnet_training_job(sagemaker_session, cpu_instance_type):
     with timeout(minutes=TRAINING_DEFAULT_TIMEOUT_MINUTES):
         script_path = os.path.join(DATA_DIR, "mxnet_mnist", "mnist_neo.py")
         data_path = os.path.join(DATA_DIR, "mxnet_mnist")
 
         mx = MXNet(
             entry_point=script_path,
             role="SageMakerRole",
-            framework_version=mxnet_full_version,
+            framework_version=NEO_MXNET_VERSION,
             py_version=PYTHON_VERSION,
             train_instance_count=1,
             train_instance_type=cpu_instance_type,
@@ -52,13 +55,10 @@ def mxnet_training_job(sagemaker_session, mxnet_full_version, cpu_instance_type)
 
 @pytest.mark.canary_quick
 @pytest.mark.regional_testing
-@pytest.mark.skip(
-    reason="This should be enabled along with the Boto SDK release for Neo API changes"
-)
 def test_attach_deploy(
     mxnet_training_job, sagemaker_session, cpu_instance_type, cpu_instance_family
 ):
-    endpoint_name = "test-mxnet-attach-deploy-{}".format(sagemaker_timestamp())
+    endpoint_name = unique_name_from_base("test-neo-attach-deploy")
 
     with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         estimator = MXNet.attach(mxnet_training_job, sagemaker_session=sagemaker_session)
@@ -77,13 +77,10 @@ def test_attach_deploy(
         predictor.predict(data)
 
 
-@pytest.mark.skip(
-    reason="This should be enabled along with the Boto SDK release for Neo API changes"
-)
 def test_deploy_model(
     mxnet_training_job, sagemaker_session, cpu_instance_type, cpu_instance_family
 ):
-    endpoint_name = "test-mxnet-deploy-model-{}".format(sagemaker_timestamp())
+    endpoint_name = unique_name_from_base("test-neo-deploy-model")
 
     with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         desc = sagemaker_session.sagemaker_client.describe_training_job(
@@ -97,14 +94,15 @@ def test_deploy_model(
             role,
             entry_point=script_path,
             py_version=PYTHON_VERSION,
+            framework_version=NEO_MXNET_VERSION,
             sagemaker_session=sagemaker_session,
         )
 
         model.compile(
             target_instance_family=cpu_instance_family,
             input_shape={"data": [1, 1, 28, 28]},
             role=role,
-            job_name="test-deploy-model-compilation-job-{}".format(int(time.time())),
+            job_name=unique_name_from_base("test-deploy-model-compilation-job"),
             output_path="/".join(model_data.split("/")[:-1]),
         )
         predictor = model.deploy(1, cpu_instance_type, endpoint_name=endpoint_name)