diff --git a/paddle/phi/kernels/gpu/full_kernel.cu b/paddle/phi/kernels/gpu/full_kernel.cu
index 79b71b95d9ee81320061832edca481d3e2ce65ef..50e57a46317e35a7fc0cf1f27d496b88ffc846d3 100644
--- a/paddle/phi/kernels/gpu/full_kernel.cu
+++ b/paddle/phi/kernels/gpu/full_kernel.cu
@@ -132,6 +132,7 @@ PD_REGISTER_KERNEL(full_like,
                    phi::FullLikeKernel,
                    float,
                    double,
+                   uint8_t,
                    int16_t,
                    int,
                    int64_t,
diff --git a/python/paddle/distributed/spawn.py b/python/paddle/distributed/spawn.py
index cea831d9d90b55d94713b9cd592665da24de7520..4adb19830522bafb99939f42f62edd1bb27c189f 100644
--- a/python/paddle/distributed/spawn.py
+++ b/python/paddle/distributed/spawn.py
@@ -452,11 +452,12 @@ def spawn(func, args=(), nprocs=-1, join=True, daemon=False, **options):
 
             def train(print_result=False):
                 # 1. initialize parallel environment
-                dist.init_parallel_env()
+                group = dist.init_parallel_env()
+                process_group = group.process_group if group else None
 
                 # 2. create data parallel layer & optimizer
                 layer = LinearNet()
-                dp_layer = paddle.DataParallel(layer)
+                dp_layer = paddle.DataParallel(layer, process_group=process_group)
 
                 loss_fn = nn.MSELoss()
                 adam = opt.Adam(
diff --git a/python/paddle/tensor/logic.py b/python/paddle/tensor/logic.py
index f11e21e65da0b727a3bed654c1bae529c21d9b9e..03d0f42d8417b82011fd1e197d3cba15f9b8257e 100755
--- a/python/paddle/tensor/logic.py
+++ b/python/paddle/tensor/logic.py
@@ -182,8 +182,8 @@ def equal(x, y, name=None):
         y = full(shape=[1], dtype=x.dtype, fill_value=y)
 
     if in_dygraph_mode():
-        axis = -1
-        return _C_ops.final_state_equal(x, y, axis)
+        default_axis = -1
+        return _C_ops.final_state_equal(x, y, default_axis)
     else:
         if _in_legacy_dygraph():
             return _C_ops.equal(x, y)
@@ -232,8 +232,8 @@ def greater_equal(x, y, name=None):
             print(result1)  # result1 = [True False True]
     """
     if in_dygraph_mode():
-        axis = -1
-        return _C_ops.final_state_greater_equal(x, y, axis)
+        default_axis = -1
+        return _C_ops.final_state_greater_equal(x, y, default_axis)
     else:
         if _in_legacy_dygraph():
             return _C_ops.greater_equal(x, y)
@@ -383,8 +383,8 @@ def less_than(x, y, name=None):
             print(result1)  # result1 = [False True False]
     """
     if in_dygraph_mode():
-        axis = -1
-        return _C_ops.final_state_less_than(x, y, axis)
+        default_axis = -1
+        return _C_ops.final_state_less_than(x, y, default_axis)
     else:
         if _in_legacy_dygraph():
             return _C_ops.less_than(x, y)
diff --git a/python/paddle/tensor/math.py b/python/paddle/tensor/math.py
index c6cbb533c4453e6495ead305877eeafc0fc15f4c..a69ecb6db4d93f1a55538a45cdef96e3ac3890d3 100755
--- a/python/paddle/tensor/math.py
+++ b/python/paddle/tensor/math.py
@@ -2668,6 +2668,7 @@ def cumsum(x, axis=None, dtype=None, name=None):
         x = cast(x, dtype)
 
     if in_dygraph_mode():
+        if axis is None: axis = -1
         return _C_ops.final_state_cumsum(x, axis, flatten, False, False)
     if _in_legacy_dygraph():
         if axis is None:
diff --git a/python/paddle/utils/code_gen/api.yaml b/python/paddle/utils/code_gen/api.yaml
index 13afa3fc868657ad447c4ca74879345e22b97954..e5cb8756da38a60eca380227e0bcbe4868292c25 100644
--- a/python/paddle/utils/code_gen/api.yaml
+++ b/python/paddle/utils/code_gen/api.yaml
@@ -419,14 +419,14 @@
     func : cumprod
   backward : cumprod_grad
 
-# cumsum
 - api : cumsum
   args : (Tensor x, int axis, bool flatten, bool exclusive, bool reverse)
-  output : Tensor
+  output : Tensor(out)
   infer_meta :
     func : CumsumInferMeta
   kernel :
     func : cumsum
+  backward : cumsum_grad
 
 - api : depthwise_conv2d_transpose
   args : (Tensor x, Tensor filter, int[] strides, int[] paddings, int[] output_padding, int[] output_size, str padding_algorithm, int groups, int[] dilations, str data_format)
diff --git a/python/paddle/utils/code_gen/backward.yaml b/python/paddle/utils/code_gen/backward.yaml
index 632636dea6dbb9fd1feafb8ff8e9d53979756eb3..875f06cecfb9bdbcb8bddeb741a758f7ccda8fa5 100644
--- a/python/paddle/utils/code_gen/backward.yaml
+++ b/python/paddle/utils/code_gen/backward.yaml
@@ -286,6 +286,15 @@
   kernel :
     func : cumprod_grad
 
+- backward_api : cumsum_grad
+  forward : cumsum(Tensor x, int axis, bool flatten, bool exclusive, bool reverse) -> Tensor(out)
+  infer_meta :
+    func : UnchangedInferMeta
+    param : [x]
+  args : (Tensor out_grad, int axis, bool flatten, bool exclusive, bool reverse)
+  output : Tensor(x_grad)
+  invoke : cumsum(out_grad, axis, flatten, exclusive, !reverse)
+
 - backward_api : depthwise_conv2d_transpose_grad
   forward : depthwise_conv2d_transpose(Tensor x, Tensor filter, int[] strides, int[] paddings, int[] output_padding, int[] output_size, str padding_algorithm, int groups, int[] dilations, str data_format) -> Tensor(out)
   args : (Tensor x, Tensor filter, Tensor out_grad, int[] strides, int[] paddings, int[] output_padding, int[] output_size, str padding_algorithm, int groups, int[] dilations, str data_format)