Merge branch 'master' into master

Author: mohamedzeidan2021

src/sagemaker/fw_utils.py

@@ -152,7 +152,6 @@
     "2.1.0",
     "2.1.2",
     "2.2.0",
-    "2.3.0",
     "2.3.1",
 ]
 

src/sagemaker/image_uri_config/pytorch-smp.json

@@ -8,7 +8,7 @@
             "2.1": "2.1.2",
             "2.2": "2.3.1",
             "2.2.0": "2.3.1",
-            "2.3": "2.4.0"
+            "2.3.1": "2.4.0"
         },
         "versions": {
             "2.0.1": {

src/sagemaker/image_uri_config/pytorch.json

@@ -1008,6 +1008,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"
@@ -1052,6 +1053,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"
@@ -2331,6 +2333,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"
@@ -2419,6 +2422,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"

src/sagemaker/image_uri_config/tensorflow.json

@@ -2140,6 +2140,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"
@@ -2181,6 +2182,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"
@@ -4354,6 +4356,7 @@
                     "us-gov-west-1": "442386744353",
                     "us-iso-east-1": "886529160074",
                     "us-isob-east-1": "094389454867",
+                    "us-isof-east-1": "303241398832",
                     "us-isof-south-1": "454834333376",
                     "us-west-1": "763104351884",
                     "us-west-2": "763104351884"

src/sagemaker/session.py

@@ -8348,7 +8348,9 @@ def _logs_for_job(  # noqa: C901 - suppress complexity warning for this method
     """
     sagemaker_client = sagemaker_session.sagemaker_client
     request_end_time = time.time() + timeout if timeout else None
-    description = sagemaker_client.describe_training_job(TrainingJobName=job_name)
+    description = _wait_until(
+        lambda: sagemaker_client.describe_training_job(TrainingJobName=job_name)
+    )
     print(secondary_training_status_message(description, None), end="")
 
     instance_count, stream_names, positions, client, log_group, dot, color_wrap = _logs_init(

tests/unit/sagemaker/image_uris/test_smp_v2.py

@@ -27,6 +27,7 @@ def test_smp_v2(load_config):
         "torch_distributed": {"enabled": True},
         "smdistributed": {"modelparallel": {"enabled": True}},
     }
+
     for processor in PROCESSORS:
         for version in VERSIONS:
             ACCOUNTS = load_config["training"]["versions"][version]["registries"]
@@ -38,6 +39,11 @@ def test_smp_v2(load_config):
                         if "2.1" in version or "2.2" in version or "2.3" in version:
                             cuda_vers = "cu121"
 
+                        if "2.3.1" == version:
+                            py_version = "py311"
+
+                        print(version, py_version)
+
                         uri = image_uris.get_training_image_uri(
                             region,
                             framework="pytorch",