feature: add support for PyTorch 1.9.0

src/sagemaker/fw_utils.py

@@ -60,7 +60,7 @@
 )
 SM_DATAPARALLEL_SUPPORTED_FRAMEWORK_VERSIONS = {
     "tensorflow": ["2.3", "2.3.1", "2.3.2", "2.4", "2.4.1"],
-    "pytorch": ["1.6", "1.6.0", "1.7", "1.7.1", "1.8", "1.8.0", "1.8.1"],
+    "pytorch": ["1.6", "1.6.0", "1.7", "1.7.1", "1.8", "1.8.0", "1.8.1", "1.9", "1.9.0"],
 }
 SMDISTRIBUTED_SUPPORTED_STRATEGIES = ["dataparallel", "modelparallel"]
 
@@ -298,7 +298,7 @@ def framework_name_from_image(image_uri):
         (tensorflow|mxnet|chainer|pytorch|scikit-learn|xgboost
         |huggingface-tensorflow|huggingface-pytorch)(?:-)?
         (scriptmode|training)?
-        :(.*)-(.*?)-(py2|py3[67]?)(?:.*)$""",
+        :(.*)-(.*?)-(py2|py3\d*)(?:.*)$""",
         re.VERBOSE,
     )
     name_match = name_pattern.match(sagemaker_match.group(9))
@@ -329,7 +329,7 @@ def framework_version_from_tag(image_tag):
     Returns:
         str: The framework version.
     """
-    tag_pattern = re.compile("^(.*)-(cpu|gpu)-(py2|py3[67]?)$")
+    tag_pattern = re.compile(r"^(.*)-(cpu|gpu)-(py2|py3\d*)$")
     tag_match = tag_pattern.match(image_tag)
     return None if tag_match is None else tag_match.group(1)
 

src/sagemaker/image_uri_config/pytorch.json

@@ -62,7 +62,8 @@
             "1.5": "1.5.0",
             "1.6": "1.6.0",
             "1.7": "1.7.1",
-            "1.8": "1.8.1"
+            "1.8": "1.8.1",
+            "1.9": "1.9.0"
         },
         "versions": {
             "0.4.0": {
@@ -433,6 +434,39 @@
                     "us-west-2": "763104351884"
                 },
                 "repository": "pytorch-inference"
+            },
+            "1.9.0": {
+                "py_versions": [
+                    "py38"
+                ],
+                "registries": {
+                    "af-south-1": "626614931356",
+                    "ap-east-1": "871362719292",
+                    "ap-northeast-1": "763104351884",
+                    "ap-northeast-2": "763104351884",
+                    "ap-northeast-3": "364406365360",
+                    "ap-south-1": "763104351884",
+                    "ap-southeast-1": "763104351884",
+                    "ap-southeast-2": "763104351884",
+                    "ca-central-1": "763104351884",
+                    "cn-north-1": "727897471807",
+                    "cn-northwest-1": "727897471807",
+                    "eu-central-1": "763104351884",
+                    "eu-north-1": "763104351884",
+                    "eu-west-1": "763104351884",
+                    "eu-west-2": "763104351884",
+                    "eu-west-3": "763104351884",
+                    "eu-south-1": "692866216735",
+                    "me-south-1": "217643126080",
+                    "sa-east-1": "763104351884",
+                    "us-east-1": "763104351884",
+                    "us-east-2": "763104351884",
+                    "us-gov-west-1": "442386744353",
+                    "us-iso-east-1": "886529160074",
+                    "us-west-1": "763104351884",
+                    "us-west-2": "763104351884"
+                },
+                "repository": "pytorch-inference"
             }
         }
     },
@@ -451,7 +485,8 @@
             "1.5": "1.5.0",
             "1.6": "1.6.0",
             "1.7": "1.7.1",
-            "1.8": "1.8.1"
+            "1.8": "1.8.1",
+            "1.9": "1.9.0"
         },
         "versions": {
             "0.4.0": {
@@ -823,6 +858,39 @@
                     "us-west-2": "763104351884"
                 },
                 "repository": "pytorch-training"
+            },
+            "1.9.0": {
+                "py_versions": [
+                    "py38"
+                ],
+                "registries": {
+                    "af-south-1": "626614931356",
+                    "ap-east-1": "871362719292",
+                    "ap-northeast-1": "763104351884",
+                    "ap-northeast-2": "763104351884",
+                    "ap-northeast-3": "364406365360",
+                    "ap-south-1": "763104351884",
+                    "ap-southeast-1": "763104351884",
+                    "ap-southeast-2": "763104351884",
+                    "ca-central-1": "763104351884",
+                    "cn-north-1": "727897471807",
+                    "cn-northwest-1": "727897471807",
+                    "eu-central-1": "763104351884",
+                    "eu-north-1": "763104351884",
+                    "eu-west-1": "763104351884",
+                    "eu-west-2": "763104351884",
+                    "eu-west-3": "763104351884",
+                    "eu-south-1": "692866216735",
+                    "me-south-1": "217643126080",
+                    "sa-east-1": "763104351884",
+                    "us-east-1": "763104351884",
+                    "us-east-2": "763104351884",
+                    "us-gov-west-1": "442386744353",
+                    "us-iso-east-1": "886529160074",
+                    "us-west-1": "763104351884",
+                    "us-west-2": "763104351884"
+                },
+                "repository": "pytorch-training"
             }
         }
     }

tests/conftest.py

@@ -184,8 +184,8 @@ def mxnet_eia_latest_py_version():
 def pytorch_training_py_version(pytorch_training_version, request):
     if Version(pytorch_training_version) < Version("1.5.0"):
         return request.param
-    elif Version(pytorch_training_version) == Version("1.7.1"):
-        return "py36"
+    elif Version(pytorch_training_version) >= Version("1.9"):
+        return "py38"
     else:
         return "py3"
 
@@ -194,8 +194,8 @@ def pytorch_training_py_version(pytorch_training_version, request):
 def pytorch_inference_py_version(pytorch_inference_version, request):
     if Version(pytorch_inference_version) < Version("1.4.0"):
         return request.param
-    elif Version(pytorch_inference_version) == Version("1.7.1"):
-        return "py36"
+    elif Version(pytorch_inference_version) >= Version("1.9"):
+        return "py38"
     else:
         return "py3"
 

tests/unit/test_fw_utils.py

@@ -653,6 +653,7 @@ def test_validate_smdataparallel_args_not_raises():
         ("ml.p3.16xlarge", "pytorch", "1.8", "py3", smdataparallel_enabled),
         ("ml.p3.16xlarge", "tensorflow", "2.4.1", "py3", smdataparallel_enabled_custom_mpi),
         ("ml.p3.16xlarge", "pytorch", "1.8.0", "py3", smdataparallel_enabled_custom_mpi),
+        ("ml.p3.16xlarge", "pytorch", "1.9.0", "py3", smdataparallel_enabled_custom_mpi),
     ]
     for instance_type, framework_name, framework_version, py_version, distribution in good_args:
         fw_utils._validate_smdataparallel_args(

tests/unit/test_processing.py

@@ -284,14 +284,20 @@ def test_sklearn_with_all_parameters_via_run_args_called_twice(
 @patch("os.path.exists", return_value=True)
 @patch("os.path.isfile", return_value=True)
 def test_pytorch_processor_with_required_parameters(
-    exists_mock, isfile_mock, botocore_resolver, sagemaker_session, pytorch_training_version
+    exists_mock,
+    isfile_mock,
+    botocore_resolver,
+    sagemaker_session,
+    pytorch_training_version,
+    pytorch_training_py_version,
 ):
     botocore_resolver.return_value.construct_endpoint.return_value = {"hostname": ECR_HOSTNAME}
 
     processor = PyTorchProcessor(
         role=ROLE,
         instance_type="ml.m4.xlarge",
         framework_version=pytorch_training_version,
+        py_version=pytorch_training_py_version,
         instance_count=1,
         sagemaker_session=sagemaker_session,
     )
@@ -302,12 +308,16 @@ def test_pytorch_processor_with_required_parameters(
 
     if version.parse(pytorch_training_version) < version.parse("1.2"):
         pytorch_image_uri = (
-            "520713654638.dkr.ecr.us-west-2.amazonaws.com/sagemaker-pytorch:{}-cpu-py3"
-        ).format(pytorch_training_version)
+            "520713654638.dkr.ecr.us-west-2.amazonaws.com/sagemaker-pytorch:{}-cpu-{}".format(
+                pytorch_training_version, pytorch_training_py_version
+            )
+        )
     else:
         pytorch_image_uri = (
-            "763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training:{}-cpu-py3"
-        ).format(pytorch_training_version)
+            "763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training:{}-cpu-{}".format(
+                pytorch_training_version, pytorch_training_py_version
+            )
+        )
 
     expected_args["app_specification"]["ImageUri"] = pytorch_image_uri