Merge branch 'kedeng/fixCrash'

Source/ComputationNetworkLib/LinearAlgebraNodes.h

@@ -749,7 +749,7 @@ class TimesNodeBase : public ComputationNode<ElemType>, public NumInputs<2>
                 }
                 else
                 {
-                    ElementTimesNode<ElemType>::BackpropToImpl(*this, inputIndex, fr, false/*allowBroadcast*/);
+                    ElementTimesNode<ElemType>::BackpropToImpl(*this, inputIndex, fr, true/*allowBroadcast*/);
                 }
                 return;
             }

bindings/python/cntk/ops/tests/linear_test.py

@@ -388,6 +388,20 @@ def test_op_transpose_times(left_operand, right_operand, device_id, precision):
     _test_binary_op(precision, device_id, times_transpose,
                     left_operand, right_operand, expected_forward, expected_backward)
 
+def test_times_transpose_sequence_param(device_id, precision):
+    dt_precision = PRECISION_TO_TYPE[precision]
+
+    from cntk import times_transpose, parameter, sequence, Value
+    dim = 5
+    num_sequences = 2
+    seq = [i for i in range(dim)]
+    identity = np.identity(dim, dtype=dt_precision)
+    input_data = Value.one_hot([seq]*num_sequences, dim, dtype=dt_precision)
+    input_var  = sequence.input_variable(shape=(dim), needs_gradient=True, dtype=dt_precision)
+    e = parameter(shape = (dim,), init = 1, dtype=dt_precision)
+    z = times_transpose(e, input_var)
+    e_grad = z.grad({input_var : input_data}, [e, input_var])
+
 def test_op_times_sparse_grad(device_id, precision):
     dt_precision = PRECISION_TO_TYPE[precision]
 
@@ -401,7 +415,7 @@ def test_op_times_sparse_grad(device_id, precision):
     e = parameter(shape = (dim, dim), init = identity, dtype=dt_precision)
     z = reshape(times_transpose(e, times(input_var, e)), dim)
     e_grad = z.grad({input_var : input_data}, [e])
-    
+
     assert np.allclose(e_grad, np.ones((dim,dim))*4)
 
 def test_op_times_reduce_sequence_axis(device_id, precision):