diff --git a/tests/python/unittest/test_module.py b/tests/python/unittest/test_module.py
index c3c7f12bbc57..876f2334fac7 100644
--- a/tests/python/unittest/test_module.py
+++ b/tests/python/unittest/test_module.py
@@ -258,7 +258,9 @@ def mean_abs(x):
                 break
     assert(mon_result_counts == [2, 2, 1, 6, 6, 4])
 
-def test_fm_module():
+def test_module_fm():
+    mx.random.seed(11)
+    rnd.seed(11)
     def fm_model(k, feature_dim, storage_type='default'):
          initializer = mx.initializer.Normal(sigma=0.01)
          x = mx.symbol.Variable("data", storage_type=storage_type)
@@ -287,8 +289,9 @@ def fm_model(k, feature_dim, storage_type='default'):
 
     num_batches = 8
     batch_size = 25
-    scipy_data = scipy.sparse.rand(num_batches * batch_size, feature_dim,
-                                   density=0.5, format='csr')
+    import scipy.sparse as sp
+    scipy_data = sp.rand(num_batches * batch_size, feature_dim,
+                         density=0.5, format='csr')
     dns_label = mx.nd.ones((num_batches * batch_size,1))
     csr_data = mx.sparse_nd.csr(scipy_data.data, scipy_data.indptr, scipy_data.indices,
                                 (num_batches * batch_size, feature_dim))
@@ -309,8 +312,7 @@ def fm_model(k, feature_dim, storage_type='default'):
     # use accuracy as the metric
     metric = mx.metric.create('MSE')
     # train 5 epoch, i.e. going over the data iter one pass
-    # TODO(haibin) test with row_sparse instead
-    storage_type_dict = {'v' : 'default'}
+    storage_type_dict = {'v' : 'row_sparse'}
 
     for epoch in range(10):
         train_iter.reset()
@@ -320,7 +322,8 @@ def fm_model(k, feature_dim, storage_type='default'):
             mod.update_metric(metric, batch.label)  # accumulate prediction accuracy
             mod.backward()                          # compute gradients
             mod.update(storage_type_dict)           # update parameters
-        print('Epoch %d, Training %s' % (epoch, metric.get()))
+        # print('Epoch %d, Training %s' % (epoch, metric.get()))
+    assert(metric.get()[1] < 0.2)
 
 if __name__ == '__main__':
     test_module_dtype()
@@ -331,4 +334,4 @@ def fm_model(k, feature_dim, storage_type='default'):
     test_module_layout()
     test_module_switch_bucket()
     test_monitor()
-    test_fm_module()
+    test_module_fm()