Support for Mlflow models that return dataframes (#3573)

Author: szczeles

servers/mlflowserver/mlflowserver/MLFlowServer.py

@@ -7,7 +7,7 @@
 
 from mlflow import pyfunc
 from seldon_core import Storage
-from seldon_core.user_model import SeldonComponent
+from seldon_core.user_model import SeldonComponent, SeldonNotImplementedError
 from typing import Dict, List, Union
 
 logger = logging.getLogger()
@@ -23,6 +23,7 @@ def __init__(self, model_uri: str, xtype: str = "ndarray"):
         self.model_uri = model_uri
         self.xtype = xtype
         self.ready = False
+        self.column_names = None
 
     def load(self):
         logger.info(f"Downloading model from {self.model_uri}")
@@ -47,6 +48,11 @@ def predict(
                 df = pd.DataFrame(data=X)
             result = self._model.predict(df)
 
+        if isinstance(result, pd.DataFrame):
+            if self.column_names is None:
+                self.column_names = result.columns.to_list()
+            result = result.to_numpy()
+
         logger.debug(f"Prediction result: {result}")
         return result
 
@@ -64,3 +70,9 @@ def init_metadata(self):
                 f"metadata file {file_path} present but does not contain valid yaml"
             )
             return {}
+
+    def class_names(self):
+        if self.column_names is not None:
+            return self.column_names
+
+        raise SeldonNotImplementedError("prediction result is not a dataframe")