diff --git a/paddle/fluid/operators/lstmp_op.h b/paddle/fluid/operators/lstmp_op.h
index 8424aa8723be8bde681313375e5380bcb96f5754..9cad0bfd0427d77419557f4b79b11f4379befd6a 100644
--- a/paddle/fluid/operators/lstmp_op.h
+++ b/paddle/fluid/operators/lstmp_op.h
@@ -225,7 +225,6 @@ class LSTMPKernel : public framework::OpKernel<T> {
         // Since the batch computing for LSTMP reorders the input sequence
         // according to their length. The initialized hidden state also needs
         // to reorder.
-        VLOG(1) << "qxz h0 used";
         ReorderInitState<DeviceContext, T>(device_ctx, *hidden_t0, order,
                                            &ordered_h0, true);
         blas.MatMul(ordered_h0, false, *weight, false, static_cast<T>(1.0),
diff --git a/python/paddle/fluid/tests/unittests/test_lstmp_op.py b/python/paddle/fluid/tests/unittests/test_lstmp_op.py
index 98252f86cce028da369eead145c584e2a7cd7fe0..299a8c9695bc1f8996c7c11056575d6aaead8f14 100644
--- a/python/paddle/fluid/tests/unittests/test_lstmp_op.py
+++ b/python/paddle/fluid/tests/unittests/test_lstmp_op.py
@@ -44,7 +44,6 @@ def lstmp(
         act_proj=None):
     def _step(x, w_r, w_rh, w_c, r_pre, c_pre, proj_clip, cell_clip, act_gate,
               act_cell, act_cand, act_proj):
-        #import pdb; pdb.set_trace()
         g = np.dot(r_pre, w_r)  # 1 x 4D
         g = g + x
         g = np.reshape(g, (1, g.size))
@@ -59,9 +58,6 @@ def lstmp(
         c = g_f * c_pre + g_i * act_cand(c)  # 1 x D
 
         def array_clip(a, clip):
-            #print('clip:{}'.format(clip))
-            #print('old' + str(a))
-
             size = np.prod(a.shape)
             new_a = np.reshape(a, (size))
             for i in range(size):
@@ -134,92 +130,17 @@ class TestLstmpOp(LstmTest.TestLstmOp):
     def reset_argument(self):
         pass
 
-    def setUp2(self):
-        self.set_argument()
-        # projection size
-        self.P = 2
-
-        self.reset_argument()
-        self.op_type = 'lstmp'
-        self.act_proj = 'identity'
-        self.use_peepholes = False
-        self.has_initial_state = True
-        self.lod = [[5]]
-
-        T = sum(self.lod[0])
-        N = len(self.lod[0])
-
-        proj_clip = 0.5
-        cell_clip = 0.0
-
-        #import pdb; pdb.set_trace()
-        x=np.array([[-0.50806344, 0.50909436], \
-  [-0.50087136, 0.4904187 ], \
-  [-0.48933774, 0.50408053], \
-  [ 0.00896523, 0.00770854], \
-  [-0.00851139,-0.01005108]])
-        wx = np.array([[ 0.2932311,  -0.8829277,   1.100133,    0.8197811,  -0.8194872,  -0.829262, 0.7708865,  -0.62339246, -0.7656475,   0.4283645,  -0.27164033, -0.3600223 ], \
-            [-0.609142,    0.25025278,  0.15731744, -0.66051376, -0.70994514,  0.8344964, -0.00551117, -0.7072167,  -0.63929003, -0.52340907, -0.8842589,   0.9531688 ]])
-        x = np.dot(x, wx)
-
-        w = np.array([[ 0.7808204, -0.7412322,  -0.9458036,  -0.01664658,  0.7930616,   0.10208707, 0.20036687, -0.16743736,  1.0295134,  -0.3118722,   0.02241168,  0.3154219 ], \
- [-0.29026014,  0.24638331, -0.5435432,   0.87635124, -0.96091515, -0.1411362, 0.58606523, -0.38996056, -0.9003789,   0.8540163,  -0.8831781,  -0.28499633]])
-
-        w_rh = np.array([[0.15685119, 0.05694652], [-0.9641068, -1.5106804],
-                         [0.3599193, 1.2540514]])
-        w_b = np.array([[
-            -0.49999997, 0.5, -0.49999997, -0.5, 0.5, 0.5, 0.49999997,
-            -0.49999997, 0.49999997, -0.5, 0.49999997, 0.5
-        ]])
-        h0 = np.array([[-1.3392334e-04, -6.8468950e-04]])
-        c0 = np.array([[4.5552300e-04, 1.3302206e-03, -3.6721351e-04]])
-        w_c = None
-        self.lod = [[5]]
-        #import pdb; pdb.set_trace()
-        r, c = lstmp(x, self.lod, h0, c0, w, w_rh, w_b, w_c, self.is_reverse,
-                     proj_clip, cell_clip, ACTIVATION[self.act_gate],
-                     ACTIVATION[self.act_cell], ACTIVATION[self.act_cand],
-                     ACTIVATION[self.act_proj])
-        self.inputs = {'Input': (x, self.lod), 'Weight': w, 'ProjWeight': w_rh}
-
-        self.inputs['Bias'] = w_b
-
-        if self.has_initial_state:
-            self.inputs['H0'] = h0
-            self.inputs['C0'] = c0
-
-        self.outputs = {
-            'Projection': (r, self.lod),
-            'Cell': (c, self.lod),
-        }
-        self.attrs = {
-            'use_peepholes': self.use_peepholes,
-            'is_reverse': self.is_reverse,
-            'proj_clip': proj_clip,
-            'cell_clip': cell_clip,
-            'gate_activation': self.act_gate,
-            'cell_activation': self.act_cell,
-            'candidate_activation': self.act_cand,
-            'proj_activation': self.act_proj
-        }
-
     def setUp(self):
         self.set_argument()
         # projection size
         self.P = 10
-        #self.D = 9
         self.act_proj = self.act_cell
 
         self.reset_argument()
         self.op_type = 'lstmp'
-        #self.use_peepholes=False
-        #self.lod=[[7]]
-        #self.act_proj='identity'
-        #self.act_proj='tanh'
 
         T = sum(self.lod[0])
         N = len(self.lod[0])
-        #np.random.seed=123
         x = np.random.normal(size=(T, 4 * self.D)).astype('float64')
         if self.has_initial_state:
             h0 = np.random.normal(size=(N, self.P)).astype('float64')
@@ -238,7 +159,6 @@ class TestLstmpOp(LstmTest.TestLstmOp):
         w_rh = np.random.normal(size=(self.D, self.P)).astype('float64')
         proj_clip = 0.1
         cell_clip = 0.1
-        #import pdb; pdb.set_trace()
         r, c = lstmp(x, self.lod, h0, c0, w, w_rh, w_b, w_c, self.is_reverse,
                      proj_clip, cell_clip, ACTIVATION[self.act_gate],
                      ACTIVATION[self.act_cell], ACTIVATION[self.act_cand],