diff --git a/examples/python/tensorflow/run.py b/examples/python/tensorflow/run.py
index 96f1c61ff..96e3cb309 100644
--- a/examples/python/tensorflow/run.py
+++ b/examples/python/tensorflow/run.py
@@ -58,12 +58,15 @@ def blocksparse_matmul_grad(op, dy):
 def run_shift():
     B, C, H, W = 1, 16, 4, 4
     R, S, F = 3, 3, 16
+    np.random.seed(2)
     a = tf.placeholder(tf.float32, shape=[C, H, W, B])
     b = tf.placeholder(tf.float32, shape=[C, F])
-    #hshift_h = np.random.randint(-R//2, R//2 + 1, size=C, dtype=np.int32)
-    #hshift_w = np.random.randint(-S//2, R//2 + 1, size=C, dtype=np.int32)
-    hshift_h = np.ones(C, dtype=np.int32)
-    hshift_w = np.ones(C, dtype=np.int32)
+    hshift_h = np.random.randint(- (R//2), R//2 + 1, size=C, dtype=np.int32)
+    hshift_w = np.random.randint(- (S//2), R//2 + 1, size=C, dtype=np.int32)
+    print(hshift_h)
+    print(hshift_w)
+    #hshift_h = np.ones(C, dtype=np.int32)
+    #hshift_w = np.ones(C, dtype=np.int32)
     c = module.shift_conv(a, b, shift_h=tf.make_tensor_proto(hshift_h), shift_w=tf.make_tensor_proto(hshift_w))
     # Reference
     ha = np.random.rand(C, H, W, B)
@@ -75,13 +78,16 @@ def run_shift():
                                     extra_feed_dict={a: ha, b: hb})
     dx_t, dx_n = grads[0]
     dw_t, dw_n = grads[1]
-    print(dx_t)
-    print(dx_n)
-    #print(np.max(dw_t - dw_n))
+    print(dw_t)
+    print(dw_n)
+    print(np.max(dw_t - dw_n))
+    #np.savetxt('diff.dat', dw_t - dw_n, fmt='%2.4f')
+    #np.savetxt('theoretical.dat', dw_t, fmt='%2.4f')
+    #np.savetxt('numerical.dat', dw_n, fmt='%2.4f')
     print(np.max(dx_t - dx_n))
-    np.savetxt('diff.dat', dx_t - dx_n, fmt='%2.4f')
-    np.savetxt('theoretical.dat', dx_t, fmt='%2.4f')
-    np.savetxt('numerical.dat', dx_n, fmt='%2.4f')
+    #np.savetxt('diff.dat', dx_t - dx_n, fmt='%2.4f')
+    #np.savetxt('theoretical.dat', dx_t, fmt='%2.4f')
+    #np.savetxt('numerical.dat', dx_n, fmt='%2.4f')
     # Run
     sess.run(tf.global_variables_initializer())
     result = sess.run([c], feed_dict = {a: ha,
diff --git a/lib/dnn/shift.cpp b/lib/dnn/shift.cpp
index 4b0662755..c7013e281 100644
--- a/lib/dnn/shift.cpp
+++ b/lib/dnn/shift.cpp
@@ -287,15 +287,15 @@ else{
 if(ty_ == BPROP){
   os << R"(
   int32 rcwhc[TM] = rxc / ABS;
-  int32 rcw[TM] = (rcwhc % AW);
+  int32 rcw[TM] = rcwhc % AW;
   int32 rchc[TM] = rcwhc / AW;
-  int32 rch[TM] = (rchc % AH);
+  int32 rch[TM] = rchc % AH;
   int1 maskh[TM] = (rch >= pad_h) && (rch < (AH - pad_h));
   int1 maskw[TM] = (rcw >= pad_w) && (rcw < (AW - pad_w));
   int1 interior[TM, TN] = maskh[:, newaxis] && maskw[:, newaxis];
   __constant__ int32* pd[TN] = delta + ryc;
   fp32* shift_pc[TM, TN] = pc + (*pd)[newaxis, :];
-  pc = interior ? shift_pc : pc;
+  pc = interior ? pc : shift_pc;
   @checkc __atomic_add(pc, C);
   )";
 }