docs

Author: phi-dbq

python/sparkdl/transformers/param.py

@@ -222,8 +222,11 @@ class HasTFHParams(Params):
 # New in sparkdl
 
 class HasOutputMapping(Params):
+    """
+    Mixin for param outputMapping: ordered list of ('outputTensorName', 'outputColName') pairs
+    """
     outputMapping = Param(Params._dummy(), "outputMapping",
-                          "Name of output tensor in signature def",
+                          "Mapping output :class:`tf.Tensor` objects to DataFrame column names",
                           typeConverter=SparkDLTypeConverters.asTensorToColumnMap)
 
     def __init__(self):
@@ -237,8 +240,11 @@ def getOutputMapping(self):
 
 
 class HasInputMapping(Params):
+    """
+    Mixin for param inputMapping: ordered list of ('inputColName', 'inputTensorName') pairs
+    """
     inputMapping = Param(Params._dummy(), "inputMapping",
-                         "Name of input tensor in signature def",
+                         "Mapping input DataFrame column names to :class:`tf.Tensor` objects",
                          typeConverter=SparkDLTypeConverters.asColumnToTensorMap)
 
     def __init__(self):
@@ -252,6 +258,9 @@ def getInputMapping(self):
 
 
 class HasTFGraph(Params):
+    """
+    Mixin for param tfGraph: the :class:`tf.Graph` object that represents a TensorFlow computation.
+    """
     tfGraph = Param(Params._dummy(), "tfGraph",
                             "TensorFlow Graph object",
                             typeConverter=SparkDLTypeConverters.toTFGraph)