Merge branch 'master' into fastapi-inprocess

Author: makungaj1

CHANGELOG.md

@@ -1,5 +1,36 @@
 # Changelog
 
+## v2.229.0 (2024-08-15)
+
+### Features
+
+ * Support for ModelBuilder In_Process Mode (1/2)
+ * Pulling in dependencies (in_process mode) using conda environment
+ * Add optional CodeArtifact login to FrameworkProcessing job script
+ * implemented security-monitoring to send metrics to CW #1510
+
+### Bug Fixes and Other Changes
+
+ * alt configs model deployment and training issues
+ * fix keras extension in integ test
+ * update image_uri_configs  08-13-2024 07:17:54 PST
+ * trn1 instance family does not support volume size
+ * Update model.py
+ * removed log statement
+ * update image_uri_configs  08-09-2024 07:18:00 PST
+ * Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance)
+ * BiasConfig type hint
+ * add model monitor image accounts for ap-southeast-5 and eu-central-2
+ * aligned UTC times with PST
+ * ensure hpt jobs inherit tags from config
+ * add JumpStart PDT and OSU regions
+ * chore(deps): bump certifi in /src/sagemaker/serve/utils
+ * Updates for DJL 0.29.0 release
+ * chore(deps): bump apache-airflow from 2.9.2 to 2.9.3 in /requirements/extras
+ * chore(deps): bump torch from 2.0.1 to 2.2.0 in /tests/data/serve_resources/mlflow/pytorch
+ * avoided printing stack trace and escaped input
+ * removing kwargs as this is breaking predictor_cls param for mode…
+
 ## v2.228.0 (2024-08-06)
 
 ### Features

VERSION

@@ -1 +1 @@
-2.228.1.dev0
+2.229.1.dev0

src/sagemaker/image_uri_config/tensorflow.json

@@ -2465,7 +2465,8 @@
             "2.11": "2.11.0",
             "2.12": "2.12.0",
             "2.13": "2.13.0",
-            "2.14": "2.14.1"
+            "2.14": "2.14.1",
+            "2.16": "2.16.2"
         },
         "versions": {
             "1.4.1": {
@@ -4407,6 +4408,47 @@
                     "us-west-2": "763104351884"
                 },
                 "repository": "tensorflow-training"
+            },
+            "2.16.2": {
+                "py_versions": [
+                    "py310"
+                ],
+                "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-south-2": "772153158452",
+                    "ap-southeast-1": "763104351884",
+                    "ap-southeast-2": "763104351884",
+                    "ap-southeast-3": "907027046896",
+                    "ap-southeast-4": "457447274322",
+                    "ca-central-1": "763104351884",
+                    "ca-west-1": "204538143572",
+                    "cn-north-1": "727897471807",
+                    "cn-northwest-1": "727897471807",
+                    "eu-central-1": "763104351884",
+                    "eu-central-2": "380420809688",
+                    "eu-north-1": "763104351884",
+                    "eu-south-1": "692866216735",
+                    "eu-south-2": "503227376785",
+                    "eu-west-1": "763104351884",
+                    "eu-west-2": "763104351884",
+                    "eu-west-3": "763104351884",
+                    "il-central-1": "780543022126",
+                    "me-central-1": "914824155844",
+                    "me-south-1": "217643126080",
+                    "sa-east-1": "763104351884",
+                    "us-east-1": "763104351884",
+                    "us-east-2": "763104351884",
+                    "us-gov-east-1": "446045086412",
+                    "us-gov-west-1": "442386744353",
+                    "us-west-1": "763104351884",
+                    "us-west-2": "763104351884"
+                },
+                "repository": "tensorflow-training"
             }
         }
     }

src/sagemaker/jumpstart/artifacts/metric_definitions.py

@@ -96,16 +96,17 @@ def _retrieve_default_training_metric_definitions(
         else []
     )
 
-    instance_specific_metric_name: str
-    for instance_specific_metric_definition in instance_specific_metric_definitions:
-        instance_specific_metric_name = instance_specific_metric_definition["Name"]
-        default_metric_definitions = list(
-            filter(
-                lambda metric_definition: metric_definition["Name"]
-                != instance_specific_metric_name,
-                default_metric_definitions,
+    if instance_specific_metric_definitions:
+        instance_specific_metric_name: str
+        for instance_specific_metric_definition in instance_specific_metric_definitions:
+            instance_specific_metric_name = instance_specific_metric_definition["Name"]
+            default_metric_definitions = list(
+                filter(
+                    lambda metric_definition: metric_definition["Name"]
+                    != instance_specific_metric_name,
+                    default_metric_definitions,
+                )
             )
-        )
-        default_metric_definitions.append(instance_specific_metric_definition)
+            default_metric_definitions.append(instance_specific_metric_definition)
 
     return default_metric_definitions

src/sagemaker/jumpstart/factory/model.py

@@ -259,6 +259,7 @@ def _add_instance_type_to_kwargs(
         sagemaker_session=kwargs.sagemaker_session,
         model_type=kwargs.model_type,
         config_name=kwargs.config_name,
+        hub_arn=kwargs.hub_arn,
     )
 
     if specs.inference_configs and kwargs.config_name not in specs.inference_configs.configs:
@@ -780,6 +781,7 @@ def _add_config_name_to_deploy_kwargs(
             sagemaker_session=temp_session,
             model_type=kwargs.model_type,
             config_name=kwargs.config_name,
+            hub_arn=kwargs.hub_arn,
         )
         default_config_name = _select_inference_config_from_training_config(
             specs=specs, training_config_name=training_config_name

src/sagemaker/jumpstart/hub/parser_utils.py

@@ -10,16 +10,20 @@
 # distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
+# pylint: skip-file
 """This module contains utilities related to SageMaker JumpStart Hub."""
 from __future__ import absolute_import
 
 import re
-from typing import Any, Dict
+from typing import Any, Dict, List, Optional
 
 
 def camel_to_snake(camel_case_string: str) -> str:
     """Converts camelCaseString or UpperCamelCaseString to snake_case_string."""
     snake_case_string = re.sub("(.)([A-Z][a-z]+)", r"\1_\2", camel_case_string)
+    if "-" in snake_case_string:
+        # remove any hyphen from the string for accurate conversion.
+        snake_case_string = snake_case_string.replace("-", "")
     return re.sub("([a-z0-9])([A-Z])", r"\1_\2", snake_case_string).lower()
 
 
@@ -29,20 +33,29 @@ def snake_to_upper_camel(snake_case_string: str) -> str:
     return upper_camel_case_string
 
 
-def walk_and_apply_json(json_obj: Dict[Any, Any], apply) -> Dict[Any, Any]:
-    """Recursively walks a json object and applies a given function to the keys."""
+def walk_and_apply_json(
+    json_obj: Dict[Any, Any], apply, stop_keys: Optional[List[str]] = ["metrics"]
+) -> Dict[Any, Any]:
+    """Recursively walks a json object and applies a given function to the keys.
+
+    stop_keys (Optional[list[str]]): List of field keys that should stop the application function.
+        Any children of these keys will not have the application function applied to them.
+    """
 
     def _walk_and_apply_json(json_obj, new):
         if isinstance(json_obj, dict) and isinstance(new, dict):
             for key, value in json_obj.items():
                 new_key = apply(key)
-                if isinstance(value, dict):
-                    new[new_key] = {}
-                    _walk_and_apply_json(value, new=new[new_key])
-                elif isinstance(value, list):
-                    new[new_key] = []
-                    for item in value:
-                        _walk_and_apply_json(item, new=new[new_key])
+                if (stop_keys and new_key not in stop_keys) or stop_keys is None:
+                    if isinstance(value, dict):
+                        new[new_key] = {}
+                        _walk_and_apply_json(value, new=new[new_key])
+                    elif isinstance(value, list):
+                        new[new_key] = []
+                        for item in value:
+                            _walk_and_apply_json(item, new=new[new_key])
+                    else:
+                        new[new_key] = value
                 else:
                     new[new_key] = value
         elif isinstance(json_obj, dict) and isinstance(new, list):
@@ -51,6 +64,8 @@ def _walk_and_apply_json(json_obj, new):
             new.update(json_obj)
         elif isinstance(json_obj, list) and isinstance(new, list):
             new.append(json_obj)
+        elif isinstance(json_obj, str) and isinstance(new, list):
+            new.append(json_obj)
         return new
 
     return _walk_and_apply_json(json_obj, new={})

src/sagemaker/jumpstart/types.py

@@ -1174,7 +1174,7 @@ def __init__(self, spec: Optional[Dict[str, Any]], is_hub_content=False):
             spec (Dict[str, Any]): Dictionary representation of training config ranking.
         """
         if is_hub_content:
-            spec = {camel_to_snake(key): val for key, val in spec.items()}
+            spec = walk_and_apply_json(spec, camel_to_snake)
         self.from_json(spec)
 
     def from_json(self, json_obj: Dict[str, Any]) -> None:
@@ -1400,7 +1400,7 @@ def from_json(self, json_obj: Dict[str, Any]) -> None:
 
         if self.training_supported:
             if self._is_hub_content:
-                self.training_ecr_uri: Optional[str] = json_obj["training_ecr_uri"]
+                self.training_ecr_uri: Optional[str] = json_obj.get("training_ecr_uri")
                 self._non_serializable_slots.append("training_ecr_specs")
             else:
                 self.training_ecr_specs: Optional[JumpStartECRSpecs] = (

src/sagemaker/jumpstart/utils.py

@@ -33,6 +33,7 @@
 
 from sagemaker.jumpstart import constants, enums
 from sagemaker.jumpstart import accessors
+from sagemaker.jumpstart.hub.parser_utils import camel_to_snake, snake_to_upper_camel
 from sagemaker.s3 import parse_s3_url
 from sagemaker.jumpstart.exceptions import (
     DeprecatedJumpStartModelError,
@@ -1103,6 +1104,17 @@ def get_jumpstart_configs(
             metadata_configs.config_rankings.get("overall").rankings if metadata_configs else []
         )
 
+    if hub_arn:
+        return (
+            {
+                config_name: metadata_configs.configs[
+                    camel_to_snake(snake_to_upper_camel(config_name))
+                ]
+                for config_name in config_names
+            }
+            if metadata_configs
+            else {}
+        )
     return (
         {config_name: metadata_configs.configs[config_name] for config_name in config_names}
         if metadata_configs

src/sagemaker/utils.py

@@ -1457,11 +1457,11 @@ def volume_size_supported(instance_type: str) -> bool:
         # Any instance type with a "d" in the instance family (i.e. c5d, p4d, etc)
         # + g5 or g6 or p5 does not support attaching an EBS volume.
         family = parts[0]
-        return (
-            "d" not in family
-            and not family.startswith("g5")
-            and not family.startswith("g6")
-            and not family.startswith("p5")
+
+        unsupported_families = ["g5", "g6", "p5", "trn1"]
+
+        return "d" not in family and not any(
+            family.startswith(prefix) for prefix in unsupported_families
         )
     except Exception as e:
         raise ValueError(f"Failed to parse instance type '{instance_type}': {str(e)}")

tests/data/tensorflow_mnist/mnist_v2.py

@@ -198,7 +198,7 @@ def main(args):
 
         if args.current_host == args.hosts[0]:
             ckpt_manager.save()
-            net.save("/opt/ml/model/1")
+            net.save("/opt/ml/model/1.keras")
 
 
 if __name__ == "__main__":