Fix:invalid component error with new metadata (#4634)

* fix: invalid component name

* tests

* format

* fix vulnerable model integ tests llama 2

* updated

* fix: training dataset location

Author: Captainia

src/sagemaker/jumpstart/estimator.py

@@ -734,7 +734,12 @@ def attach(
 
         model_version = model_version or "*"
 
-        additional_kwargs = {"model_id": model_id, "model_version": model_version}
+        additional_kwargs = {
+            "model_id": model_id,
+            "model_version": model_version,
+            "tolerate_vulnerable_model": True,  # model is already trained
+            "tolerate_deprecated_model": True,  # model is already trained
+        }
 
         model_specs = verify_model_region_and_return_specs(
             model_id=model_id,

src/sagemaker/jumpstart/types.py

@@ -1064,9 +1064,8 @@ def from_json(self, json_obj: Dict[str, Any]) -> None:
                 Dictionary representation of the config component.
         """
         for field in json_obj.keys():
-            if field not in self.__slots__:
-                raise ValueError(f"Invalid component field: {field}")
-            setattr(self, field, json_obj[field])
+            if field in self.__slots__:
+                setattr(self, field, json_obj[field])
 
 
 class JumpStartMetadataConfig(JumpStartDataHolderType):

tests/integ/sagemaker/jumpstart/constants.py

@@ -48,6 +48,7 @@ def _to_s3_path(filename: str, s3_prefix: Optional[str]) -> str:
     ("meta-textgeneration-llama-2-7b", "*"): ("training-datasets/sec_amazon/"),
     ("meta-textgeneration-llama-2-7b", "2.*"): ("training-datasets/sec_amazon/"),
     ("meta-textgeneration-llama-2-7b", "3.*"): ("training-datasets/sec_amazon/"),
+    ("meta-textgeneration-llama-2-7b", "4.*"): ("training-datasets/sec_amazon/"),
     ("meta-textgenerationneuron-llama-2-7b", "*"): ("training-datasets/sec_amazon/"),
 }
 

tests/integ/sagemaker/jumpstart/estimator/test_jumpstart_estimator.py

@@ -140,7 +140,7 @@ def test_gated_model_training_v1(setup):
 def test_gated_model_training_v2(setup):
 
     model_id = "meta-textgeneration-llama-2-7b"
-    model_version = "3.*"  # model artifacts retrieved from jumpstart-private-cache-* buckets
+    model_version = "4.*"  # model artifacts retrieved from jumpstart-private-cache-* buckets
 
     estimator = JumpStartEstimator(
         model_id=model_id,
@@ -150,6 +150,7 @@ def test_gated_model_training_v2(setup):
         tags=[{"Key": JUMPSTART_TAG, "Value": os.environ[ENV_VAR_JUMPSTART_SDK_TEST_SUITE_ID]}],
         environment={"accept_eula": "true"},
         max_run=259200,  # avoid exceeding resource limits
+        tolerate_vulnerable_model=True,  # tolerate old version of model
     )
 
     # uses ml.g5.12xlarge instance

tests/unit/sagemaker/jumpstart/estimator/test_estimator.py

@@ -1010,6 +1010,8 @@ def test_jumpstart_estimator_attach_eula_model(
                 "model_id": "gemma-model",
                 "model_version": "*",
                 "environment": {"accept_eula": "true"},
+                "tolerate_vulnerable_model": True,
+                "tolerate_deprecated_model": True,
             },
         )
 
@@ -1053,6 +1055,8 @@ def test_jumpstart_estimator_attach_no_model_id_happy_case(
             additional_kwargs={
                 "model_id": "js-trainable-model-prepacked",
                 "model_version": "1.0.0",
+                "tolerate_vulnerable_model": True,
+                "tolerate_deprecated_model": True,
             },
         )
 

tests/unit/sagemaker/jumpstart/test_types.py

@@ -1052,6 +1052,14 @@ def test_inference_configs_parsing():
     )
     assert list(config.config_components.keys()) == ["neuron-inference"]
 
+    spec = {
+        **BASE_SPEC,
+        **INFERENCE_CONFIGS,
+        **INFERENCE_CONFIG_RANKINGS,
+        "unrecognized-field": "blah",  # New fields in base metadata fields should be ignored
+    }
+    specs1 = JumpStartModelSpecs(spec)
+
 
 def test_set_inference_configs():
     spec = {**BASE_SPEC, **INFERENCE_CONFIGS, **INFERENCE_CONFIG_RANKINGS}