Python3 unit tests (apache#39)

* change xrange to range for python3 compatiblity"


* remove more xrange from tests

python/mxnet/_ctypes/ndarray.py

@@ -178,7 +178,7 @@ def %s(%s):
     if original_output is not None:
         return original_output
     ret_list = []
-    for i in xrange(num_output.value):
+    for i in range(num_output.value):
         storage_type = ctypes.c_int(0)
         check_call(_LIB.MXNDArrayGetStorageType(ctypes.cast(output_vars[i], NDArrayHandle),
                                                 ctypes.byref(storage_type)))

python/mxnet/executor.py

@@ -93,7 +93,7 @@ def _get_outputs(self):
                                           ctypes.byref(out_size), ctypes.byref(handles)))
         num_output = out_size.value
         outputs = []
-        for i in xrange(num_output):
+        for i in range(num_output):
             storage_type = ctypes.c_int(0)
             check_call(_LIB.MXNDArrayGetStorageType(ctypes.cast(handles[i], NDArrayHandle),
                                                     ctypes.byref(storage_type)))

python/mxnet/sparse_ndarray.py

@@ -329,7 +329,7 @@ def aux_types(self):
         '''
         aux_types = []
         num_aux = self._num_aux
-        for i in xrange(num_aux):
+        for i in range(num_aux):
             aux_types.append(self._aux_type(i))
         return aux_types
 

tests/python/unittest/test_optimizer.py

@@ -184,7 +184,7 @@ def update(self, index, weight, grad, state):
         num_rows = weight.shape[0]
         if self.momentum == 0.0:
             # Update on a per row basis, skip all-zero rows
-            for row in xrange(num_rows):
+            for row in range(num_rows):
                 grad_row = grad[row].asnumpy()
                 all_zeros = mx.test_utils.almost_equal(grad_row, np.zeros_like(grad_row))
                 if all_zeros:
@@ -197,7 +197,7 @@ def update(self, index, weight, grad, state):
                     weight[row] = (1 - lr*wd)*weight[row] - lr*self.rescale_grad*grad[row]
         else:
             mom = state
-            for row in xrange(num_rows):
+            for row in range(num_rows):
               grad_row = grad[row].asnumpy()
               all_zeros = mx.test_utils.almost_equal(grad_row, np.zeros_like(grad_row))
               if all_zeros:

tests/python/unittest/test_sparse_ndarray.py

@@ -25,7 +25,7 @@ def test_sparse_nd_elemwise_add():
     num_repeats = 10
     g = lambda x,y: x + y
     op = mx.nd.elemwise_add
-    for i in xrange(num_repeats):
+    for i in range(num_repeats):
         shape = [(rnd.randint(1, 10),rnd.randint(1, 10))] * 2
         check_sparse_nd_elemwise_binary(shape, ['default_storage'] * 2, op, g)
         check_sparse_nd_elemwise_binary(shape, ['default_storage', 'row_sparse'], op, g)
@@ -36,7 +36,7 @@ def test_sparse_nd_elementwise_fallback():
     num_repeats = 10
     g = lambda x,y: x + y
     op = mx.nd.add_n
-    for i in xrange(num_repeats):
+    for i in range(num_repeats):
         shape = [(rnd.randint(1, 10), rnd.randint(1, 10))] * 2
         check_sparse_nd_elemwise_binary(shape, ['default_storage'] * 2, op, g)
         check_sparse_nd_elemwise_binary(shape, ['default_storage', 'row_sparse'], op, g)

tests/python/unittest/test_sparse_operator.py

@@ -59,7 +59,7 @@ def test_elemwise_add_ex_multiple_stages():
     check_symbolic_forward(test, {'sp_data1':sp_nd1, 'sp_data2':sp_nd2,
                           'ds_data':ds_nd}, [sp_np1 + sp_np2 + ds_np])
 
-    arr_grads = [mx.nd.zeros(shape) for i in xrange(3)]
+    arr_grads = [mx.nd.zeros(shape) for i in range(3)]
     exec_test = test.bind(default_context(), args={'sp_data1':sp_nd1, 'sp_data2':sp_nd2,
                           'ds_data':ds_nd}, args_grad=arr_grads)
     exec_test.forward(is_train=True)