diff --git a/paddle/fluid/operators/sequence_ops/sequence_mask_op.h b/paddle/fluid/operators/sequence_ops/sequence_mask_op.h
index d541f712a5d670c289179c4807b5c6c8965503ef..d69ebafb017ec290b66f55d45b9a96691fd9b76d 100644
--- a/paddle/fluid/operators/sequence_ops/sequence_mask_op.h
+++ b/paddle/fluid/operators/sequence_ops/sequence_mask_op.h
@@ -106,18 +106,23 @@ class SequenceMaskKernel : public framework::OpKernel<Tx> {
 
     auto *x_data = x->data<Tx>();
     auto x_numel = x->numel();
+
     if (maxlen < 0) {
+      if (x_numel == 0) {
+        maxlen = 0;
+      } else {
 #if defined(__NVCC__) || defined(__HIPCC__)
-      VLOG(10)
-          << "SequenceMaskOp on GPU may be slow when maxlen is not provided.";
-      maxlen = static_cast<int>(
-          thrust::reduce(thrust::device_pointer_cast(x_data),
-                         thrust::device_pointer_cast(x_data) + x_numel,
-                         static_cast<Tx>(0),
-                         thrust::maximum<Tx>()));
+        VLOG(10)
+            << "SequenceMaskOp on GPU may be slow when maxlen is not provided.";
+        maxlen = static_cast<int>(
+            thrust::reduce(thrust::device_pointer_cast(x_data),
+                           thrust::device_pointer_cast(x_data) + x_numel,
+                           static_cast<Tx>(0),
+                           thrust::maximum<Tx>()));
 #else
-      maxlen = static_cast<int>(*std::max_element(x_data, x_data + x_numel));
+        maxlen = static_cast<int>(*std::max_element(x_data, x_data + x_numel));
 #endif
+      }
       auto y_dim = phi::vectorize<int>(x->dims());
       y_dim.push_back(maxlen);
       y->Resize(phi::make_ddim(y_dim));
diff --git a/paddle/fluid/operators/sequence_ops/sequence_mask_op_npu.cc b/paddle/fluid/operators/sequence_ops/sequence_mask_op_npu.cc
index f3b18676abe56440b67afb124172cafb7f40e74c..200ec54ac200edc97d653abee329d2a330dd88f8 100644
--- a/paddle/fluid/operators/sequence_ops/sequence_mask_op_npu.cc
+++ b/paddle/fluid/operators/sequence_ops/sequence_mask_op_npu.cc
@@ -48,10 +48,14 @@ class SequenceMaskNPUKernel : public framework::OpKernel<T> {
 
     if (maxlen < 0) {
       auto x_numel = x->numel();
-      std::vector<T> x_vec;
-      framework::TensorToVector(*x, dev_ctx, &x_vec);
-      auto x_data = x_vec.data();
-      maxlen = static_cast<int>(*std::max_element(x_data, x_data + x_numel));
+      if (x_numel == 0) {
+        maxlen = 0;
+      } else {
+        std::vector<T> x_vec;
+        framework::TensorToVector(*x, dev_ctx, &x_vec);
+        auto x_data = x_vec.data();
+        maxlen = static_cast<int>(*std::max_element(x_data, x_data + x_numel));
+      }
     }
     auto y_dim = phi::vectorize<int>(x->dims());
     y_dim.push_back(maxlen);
diff --git a/python/paddle/fluid/tests/unittests/sequence/test_sequence_mask.py b/python/paddle/fluid/tests/unittests/sequence/test_sequence_mask.py
index b7c3cc0e4dd3aafa1d26ec053eb2df41af2fa3c0..c8b210625571961da053a2ef3bd7bca4b702211d 100644
--- a/python/paddle/fluid/tests/unittests/sequence/test_sequence_mask.py
+++ b/python/paddle/fluid/tests/unittests/sequence/test_sequence_mask.py
@@ -17,6 +17,7 @@ import unittest
 
 import numpy as np
 
+import paddle
 import paddle.fluid as fluid
 from paddle.fluid.framework import (
     Program,
@@ -171,5 +172,14 @@ class TestSequenceMaskOpError(unittest.TestCase):
             self.assertRaises(TypeError, test_Variable)
 
 
+class TestSequenceMaskWithEmptyTensor(unittest.TestCase):
+    def test_empty(self):
+        paddle.disable_static()
+        lengths = paddle.to_tensor(np.array([], dtype=np.int64))
+        mask = paddle.nn.functional.sequence_mask(lengths)
+        self.assertEqual(list(mask.shape), [0, 0])
+        paddle.enable_static()
+
+
 if __name__ == '__main__':
     unittest.main()