Better parameter checks

LanguageBindings/Python/cntk/examples/LogReg/test1.py

@@ -4,8 +4,8 @@
 from cntk import *
 
 if (__name__ == "__main__"):
-    x = Input(2)
-    y = Input(3)
+    x = Input(2, var_name='x')
+    y = Input(3, var_name='y')
     w = LearnableParameter(3, 2)
     b = LearnableParameter(3, 1)
     t = Times(w, x)
@@ -15,17 +15,17 @@
     ec.tag = 'criterion'
 
     reader = UCIFastReader(
-        "Train-3Classes.txt", "2", "0", "1", "2", "3", "SimpleMapping-3Classes.txt")
-    #reader.add_input(x, 0, 2)
-    #reader.add_input(y, 2, 1)
+        "Train-3Classes.txt", "y", "1", "2", "3", "SimpleMapping-3Classes.txt")
+    reader.add_input('x', 0, 2)
+    reader.add_input('y', 2, 1)
 
     my_sgd = SGD(
         epoch_size=0, minibatch_size=25, learning_ratesPerMB=0.1, max_epochs=3)
 
     with Context('demo', optimizer=my_sgd, root_node= ec, clean_up=False) as ctx:
         input_map = {x: (reader, (0, 2)), y: (reader, (2, 1))}
-        ctx.train(input_map)
+        ctx.train(reader)
 
         #import ipdb;ipdb.set_trace()
         result = ctx.eval(out, input_map)
-        print(result)
+        print(result[:3])

LanguageBindings/Python/cntk/graph.py

@@ -1,11 +1,14 @@
-
 class ComputationNode(object):
     '''
     Base class for all nodes and operators. Provides a NumPy-like interface
     with operators that are converted to CNTK operators.
     '''
 
     def __init__(self, name, params=None, var_name=None):
+        if not isinstance(name, str):
+            raise ValueError("Parameter 'name' has to be a string and not '%s'"%type(name))
+        if var_name is not None and not isinstance(var_name, str):
+            raise ValueError("Parameter 'var_name' has to be a string and not '%s'"%type(var_name))
         self.name = name
         self.params = params
         self.var_name = var_name
@@ -100,7 +103,7 @@ def _param_to_brainscript(self, p_name, p_value):
                 p_value = ":".join(v for v in p_value)
             else:
                 raise ValueError('Sequence initialization is only allowed for' +
-                                 ' parameter "dims" and not "%s"' % p_name)
+                                 ' parameters dims and not "%s"' % p_name)
         else:
             p_value = str(p_value)
 
@@ -116,16 +119,26 @@ def _to_description_unroll(self, desc, unrolled_nodes, inputs, node_counter=0):
         if self.params:
             for p_name in self.params:
                 p_value = self.__dict__[p_name]
-                if hasattr(p_value, '_to_description') and p_name:
-                    if p_value in unrolled_nodes:
-                        # we have seen this node already, so just retrieve its
-                        # name
-                        child_var = unrolled_nodes[p_value]
+                if hasattr(p_value, '_to_description') and p_name or \
+                        p_name == 'inputs':
+                        # TODO this is under the assumption that RowStack's
+                        # inputs parameter gets a tuple of inputs
+
+                    if p_name == 'inputs':
+                        inputs = p_value
                     else:
-                        child_var, node_counter, child_desc = p_value._to_description_unroll(
-                            desc, unrolled_nodes, inputs, node_counter)
-                        unrolled_nodes[p_value] = child_var
-                    param_variable_names.append(child_var)
+                        inputs = [p_value]
+
+                    for p_value in inputs:
+                        if p_value in unrolled_nodes:
+                            # we have seen this node already, so just retrieve its
+                            # name
+                            child_var = unrolled_nodes[p_value]
+                        else:
+                            child_var, node_counter, child_desc = p_value._to_description_unroll(
+                                desc, unrolled_nodes, inputs, node_counter)
+                            unrolled_nodes[p_value] = child_var
+                        param_variable_names.append(child_var)
                 else:
                     param_variable_names.append(
                         self._param_to_brainscript(p_name, p_value))