fix: update output schema for regression task predictions to ensure correct data type

Author: Creylay

DashAI/back/job/predict_job.py

@@ -12,6 +12,7 @@
 from DashAI.back.job.base_job import BaseJob, JobError
 from DashAI.back.models.base_model import BaseModel
 from DashAI.back.tasks.base_task import BaseTask
+from DashAI.back.tasks.regression_task import RegressionTask
 
 if TYPE_CHECKING:
     from sqlalchemy.orm import sessionmaker
@@ -482,6 +483,14 @@ def run(
                     if key in model_session.input_columns + model_session.output_columns
                 }
 
+                # Regression models predict continuous values regardless of
+                # the training target's original dtype (e.g. a target column
+                # that happened to hold only integer-looking values), so the
+                # output column's saved schema must reflect that instead of
+                # inheriting the training dataset's type.
+                if isinstance(task, RegressionTask):
+                    filtered_schema[output_col] = {"type": "Float", "dtype": "float64"}
+
                 # Store num of rows, columns, and column names
                 dataset_with_prediction.compute_base_metadata()