googleapis
diff --git a/Collapse file
‎tests/unit/vertexai/test_evaluation.py‎
Copy file name to clipboardExpand all lines: tests/unit/vertexai/test_evaluation.py
+57-4Lines changed: 57 additions & 4 deletions b/Collapse file
‎tests/unit/vertexai/test_evaluation.py‎
Copy file name to clipboardExpand all lines: tests/unit/vertexai/test_evaluation.py
+57-4Lines changed: 57 additions & 4 deletions
diff --git a/Collapse file
‎vertexai/preview/evaluation/_eval_tasks.py‎
Copy file name to clipboardExpand all lines: vertexai/preview/evaluation/_eval_tasks.py
+8-3Lines changed: 8 additions & 3 deletions b/Collapse file
‎vertexai/preview/evaluation/_eval_tasks.py‎
Copy file name to clipboardExpand all lines: vertexai/preview/evaluation/_eval_tasks.py
+8-3Lines changed: 8 additions & 3 deletions
@@ -28,11 +28,13 @@
 )
 from vertexai import generative_models
 from vertexai.preview import evaluation
+from vertexai.preview.evaluation import _base as eval_base
+from vertexai.preview.evaluation import _evaluation
 from vertexai.preview.evaluation import utils
+import numpy as np
 import pandas as pd
 import pytest
 
-
 _TEST_PROJECT = "test-project"
 _TEST_LOCATION = "us-central1"
 _TEST_METRICS = (
@@ -78,6 +80,7 @@
 text,text,text\n
 """
 
+_TEST_EXPERIMENT = "test-experiment"
 
 _MOCK_EXACT_MATCH_RESULT = (
     gapic_evaluation_service_types.EvaluateInstancesResponse(
@@ -135,6 +138,19 @@
         ]
     }
 )
+MOCK_EVAL_RESULT = eval_base.EvalResult(
+    summary_metrics={
+        "row_count": 1,
+        "mock_metric/mean": 1.0,
+        "mock_metric/std": np.nan,
+    },
+    metrics_table=pd.DataFrame(
+        {
+            "response": ["test"],
+            "mock_metric": [1.0],
+        }
+    ),
+)
 
 
 @pytest.fixture
@@ -163,23 +179,22 @@ def teardown_method(self):
         initializer.global_pool.shutdown(wait=True)
 
     def test_create_eval_task(self):
-        test_experiment = "test_experiment_name"
         test_content_column_name = "test_content_column_name"
         test_reference_column_name = "test_reference_column_name"
         test_response_column_name = "test_response_column_name"
 
         test_eval_task = evaluation.EvalTask(
             dataset=_TEST_EVAL_DATASET,
             metrics=_TEST_METRICS,
-            experiment=test_experiment,
+            experiment=_TEST_EXPERIMENT,
             content_column_name=test_content_column_name,
             reference_column_name=test_reference_column_name,
             response_column_name=test_response_column_name,
         )
 
         assert test_eval_task.dataset.equals(_TEST_EVAL_DATASET)
         assert test_eval_task.metrics == _TEST_METRICS
-        assert test_eval_task.experiment == test_experiment
+        assert test_eval_task.experiment == _TEST_EXPERIMENT
         assert test_eval_task.content_column_name == test_content_column_name
         assert test_eval_task.reference_column_name == test_reference_column_name
         assert test_eval_task.response_column_name == test_response_column_name
@@ -470,6 +485,44 @@ def test_compute_pairwise_metrics_without_inference(self, api_transport):
             == 0.5
         )
 
+    def test_eval_result_experiment_run_logging(self):
+        test_eval_task = evaluation.EvalTask(
+            dataset=_TEST_EVAL_DATASET,
+            metrics=_TEST_METRICS,
+            experiment=_TEST_EXPERIMENT,
+        )
+
+        with mock.patch.multiple(
+            metadata._experiment_tracker,
+            _experiment=mock.MagicMock(name=_TEST_EXPERIMENT),
+            _experiment_run=None,
+            set_experiment=mock.DEFAULT,
+            reset=mock.DEFAULT,
+        ):
+            with mock.patch.multiple(
+                vertexai.preview,
+                start_run=mock.MagicMock(),
+                log_params=mock.DEFAULT,
+                log_metrics=mock.DEFAULT,
+            ) as mock_metadata:
+                with mock.patch.object(
+                    target=_evaluation,
+                    attribute="evaluate",
+                    side_effect=[MOCK_EVAL_RESULT],
+                ):
+                    test_result = test_eval_task.evaluate()
+
+        assert test_result.summary_metrics["row_count"] == 1
+        assert test_result.summary_metrics["mock_metric/mean"] == 1.0
+        assert test_result.summary_metrics["mock_metric/std"] == "NaN"
+        mock_metadata["log_metrics"].assert_called_once_with(
+            {
+                "row_count": 1,
+                "mock_metric/mean": 1.0,
+                "mock_metric/std": "NaN",
+            }
+        )
+
 
 @pytest.mark.usefixtures("google_auth_mock")
 class TestEvaluationErrors:
 
@@ -28,6 +28,7 @@
 from vertexai.preview.evaluation.metrics import (
     _base as metrics_base,
 )
+import numpy as np
 
 if TYPE_CHECKING:
     import pandas as pd
@@ -284,9 +285,14 @@ def _evaluate_with_experiment(
                 reference_column_name=self.reference_column_name,
                 response_column_name=response_column_name,
             )
+
+            eval_result.summary_metrics = {
+                k: ("NaN" if isinstance(v, float) and np.isnan(v) else v)
+                for k, v in eval_result.summary_metrics.items()
+            }
             try:
                 vertexai.preview.log_metrics(eval_result.summary_metrics)
-            except (ValueError, TypeError, exceptions.InvalidArgument) as e:
+            except (TypeError, exceptions.InvalidArgument) as e:
                 _LOGGER.warning(f"Experiment metrics logging failed: {str(e)}")
         return eval_result
 
@@ -366,8 +372,7 @@ def _validate_experiment_run(self) -> None:
         if metadata._experiment_tracker.experiment_run:
             raise ValueError(
                 "Experiment run already exists. Please specify the name of the"
-                " experiment run to assign current session with in this evaluate"
-                " method."
+                " experiment run to assign current session within this evaluation."
             )
 
     def _log_eval_experiment_param(