diff --git a/.github/workflows/badge.yml b/.github/workflows/badge.yml
index 3030f9c37e..bbf84e6b7d 100644
--- a/.github/workflows/badge.yml
+++ b/.github/workflows/badge.yml
@@ -3,8 +3,6 @@ name: Badge Creation
 on:
   push:
     branches: [ main, 'badge/*', 'doc/*' ]
-  pull_request:
-    branches: [ main, 'badge/*', 'doc/*' ]
 
 jobs:
   update-badges:
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 2f6bc063cb..1cb9facbfc 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -2,7 +2,6 @@ name: Code Test
 
 on:
   - push
-  - pull_request
 
 jobs:
   unittest:
@@ -18,27 +17,31 @@ jobs:
           - '3.7'
           - '3.8'
           - '3.9'
+        numpy-version:
+          - '1.18.0'
+          - '1.20.0'
+          - '1.22.0'
         torch-version:
+          - '1.2.0'
           - '1.4.0'
-          - '1.5.0'
           - '1.6.0'
-          - '1.7.0'
           - '1.8.0'
-          - '1.9.0'
           - '1.10.0'
         exclude:
-          - os: 'ubuntu-18.04'
-            python-version: '3.9'
+          - python-version: '3.6'
+            numpy-version: '1.20.0'
+          - python-version: '3.6'
+            numpy-version: '1.22.0'
+          - python-version: '3.7'
+            numpy-version: '1.22.0'
+          - python-version: '3.8'
+            torch-version: '1.2.0'
+          - python-version: '3.9'
+            torch-version: '1.2.0'
+          - python-version: '3.9'
             torch-version: '1.4.0'
-          - os: 'ubuntu-18.04'
-            python-version: '3.9'
-            torch-version: '1.5.0'
-          - os: 'ubuntu-18.04'
-            python-version: '3.9'
+          - python-version: '3.9'
             torch-version: '1.6.0'
-          - os: 'ubuntu-18.04'
-            python-version: '3.9'
-            torch-version: '1.7.0'
 
     steps:
       - name: Checkout code
@@ -60,6 +63,14 @@ jobs:
         run: |
           python -m pip install --upgrade pip
           pip install --upgrade flake8 setuptools wheel twine
+      - name: Install latest numpy
+        if: ${{ matrix.numpy-version == 'latest' }}
+        run: |
+          pip install 'numpy'
+      - name: Install numpy v${{ matrix.numpy-version }}
+        if: ${{ matrix.numpy-version != 'latest' }}
+        run: |
+          pip install 'numpy==${{ matrix.numpy-version }}'
       - name: Install latest pytorch
         if: ${{ matrix.torch-version == 'latest' }}
         run: |
diff --git a/docs/source/api_doc/numpy/funcs.rst.py b/docs/source/api_doc/numpy/funcs.rst.py
index f8d9f4bec5..2ab8bee53c 100644
--- a/docs/source/api_doc/numpy/funcs.rst.py
+++ b/docs/source/api_doc/numpy/funcs.rst.py
@@ -73,7 +73,7 @@ def _doc_process(doc: str) -> str:
                         print_title(f"Description From Numpy v{_short_version}", levelc='-', file=p_func)
                         current_module(np.__name__, file=p_func)
 
-                        _origin_doc = _doc_process(_origin.__doc__ or "")
+                        _origin_doc = _doc_process(_origin.__doc__ or "").lstrip()
                         _doc_lines = _origin_doc.splitlines()
                         _first_line, _other_lines = _doc_lines[0], _doc_lines[1:]
                         if _first_line.strip():
diff --git a/requirements.txt b/requirements.txt
index 40780da3dd..6015ec5bc1 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,4 +1,4 @@
-treevalue>=1.2.0
+treevalue>=1.3.0
 torch>=1.1.0,<=1.10.0
 hbutils>=0.0.1
 numpy
diff --git a/test/numpy/test_array.py b/test/numpy/test_array.py
index d4aa0ef039..6a940d0edc 100644
--- a/test/numpy/test_array.py
+++ b/test/numpy/test_array.py
@@ -1,7 +1,9 @@
 import numpy as np
 import pytest
+import torch
 
 import treetensor.numpy as tnp
+import treetensor.torch as ttorch
 from treetensor.common import Object
 
 
@@ -233,3 +235,22 @@ def test_tolist(self):
                 'd': [0, 0, 0.0],
             }
         })
+
+    def test_tensor(self):
+        assert ttorch.isclose(self._DEMO_1.tensor().double(), ttorch.Tensor({
+            'a': ttorch.Tensor([[1, 2, 3], [4, 5, 6]]),
+            'b': ttorch.Tensor([1, 3, 5, 7]),
+            'x': {
+                'c': ttorch.Tensor([[11], [23]]),
+                'd': ttorch.Tensor([3, 9, 11.0])
+            }
+        }).double()).all()
+
+        assert (self._DEMO_1.tensor(dtype=torch.float64) == ttorch.Tensor({
+            'a': ttorch.Tensor([[1, 2, 3], [4, 5, 6]], dtype=torch.float64),
+            'b': ttorch.Tensor([1, 3, 5, 7], dtype=torch.float64),
+            'x': {
+                'c': ttorch.Tensor([[11], [23]], dtype=torch.float64),
+                'd': ttorch.Tensor([3, 9, 11.0], dtype=torch.float64),
+            }
+        })).all()
diff --git a/test/numpy/test_funcs.py b/test/numpy/test_funcs.py
index 6330922940..230b0b5441 100644
--- a/test/numpy/test_funcs.py
+++ b/test/numpy/test_funcs.py
@@ -127,3 +127,97 @@ def test_array_equal(self):
                 'd': True,
             }
         })
+
+    def test_zeros(self):
+        zs = tnp.zeros((2, 3))
+        assert isinstance(zs, np.ndarray)
+        assert np.allclose(zs, np.zeros((2, 3)))
+
+        zs = tnp.zeros({'a': (2, 3), 'c': {'x': (3, 4)}})
+        assert tnp.allclose(zs, tnp.ndarray({
+            'a': np.zeros((2, 3)),
+            'c': {'x': np.zeros((3, 4))}
+        }))
+
+    def test_ones(self):
+        zs = tnp.ones((2, 3))
+        assert isinstance(zs, np.ndarray)
+        assert np.allclose(zs, np.ones((2, 3)))
+
+        zs = tnp.ones({'a': (2, 3), 'c': {'x': (3, 4)}})
+        assert tnp.allclose(zs, tnp.ndarray({
+            'a': np.ones((2, 3)),
+            'c': {'x': np.zeros((3, 4))}
+        }))
+
+    def test_stack(self):
+        a = np.array([1, 2, 3])
+        b = np.array([2, 3, 4])
+        nd = tnp.stack((a, b))
+        assert isinstance(nd, np.ndarray)
+        assert np.allclose(nd, np.array([[1, 2, 3],
+                                         [2, 3, 4]]))
+
+        a = tnp.array({
+            'a': [1, 2, 3],
+            'c': {'x': [11, 22, 33]},
+        })
+        b = tnp.array({
+            'a': [2, 3, 4],
+            'c': {'x': [22, 33, 44]},
+        })
+        nd = tnp.stack((a, b))
+        assert tnp.allclose(nd, tnp.array({
+            'a': [[1, 2, 3], [2, 3, 4]],
+            'c': {'x': [[11, 22, 33], [22, 33, 44]]},
+        }))
+
+    def test_concatenate(self):
+        a = np.array([[1, 2], [3, 4]])
+        b = np.array([[5, 6]])
+        nd = tnp.concatenate((a, b), axis=0)
+        assert isinstance(nd, np.ndarray)
+        assert np.allclose(nd, np.array([[1, 2],
+                                         [3, 4],
+                                         [5, 6]]))
+
+        a = tnp.array({
+            'a': [[1, 2], [3, 4]],
+            'c': {'x': [[11, 22], [33, 44]]},
+        })
+        b = tnp.array({
+            'a': [[5, 6]],
+            'c': {'x': [[55, 66]]},
+        })
+        nd = tnp.concatenate((a, b), axis=0)
+        assert tnp.allclose(nd, tnp.array({
+            'a': [[1, 2], [3, 4], [5, 6]],
+            'c': {'x': [[11, 22], [33, 44], [55, 66]]},
+        }))
+
+    def test_split(self):
+        x = np.arange(9.0)
+        ns = tnp.split(x, 3)
+        assert len(ns) == 3
+        assert isinstance(ns[0], np.ndarray)
+        assert np.allclose(ns[0], np.array([0.0, 1.0, 2.0]))
+        assert isinstance(ns[1], np.ndarray)
+        assert np.allclose(ns[1], np.array([3.0, 4.0, 5.0]))
+        assert isinstance(ns[2], np.ndarray)
+        assert np.allclose(ns[2], np.array([6.0, 7.0, 8.0]))
+
+        xx = tnp.arange(tnp.ndarray({'a': 9.0, 'c': {'x': 18.0}}))
+        ns = tnp.split(xx, 3)
+        assert len(ns) == 3
+        assert tnp.allclose(ns[0], tnp.array({
+            'a': [0.0, 1.0, 2.0],
+            'c': {'x': [0.0, 1.0, 2.0, 3.0, 4.0, 5.0]},
+        }))
+        assert tnp.allclose(ns[1], tnp.array({
+            'a': [3.0, 4.0, 5.0],
+            'c': {'x': [6.0, 7.0, 8.0, 9.0, 10.0, 11.0]},
+        }))
+        assert tnp.allclose(ns[2], tnp.array({
+            'a': [6.0, 7.0, 8.0],
+            'c': {'x': [12.0, 13.0, 14.0, 15.0, 16.0, 17.0]},
+        }))
diff --git a/test/torch/tensor/test_reduction.py b/test/torch/tensor/test_reduction.py
index 64db678521..7b28c3b124 100644
--- a/test/torch/tensor/test_reduction.py
+++ b/test/torch/tensor/test_reduction.py
@@ -4,6 +4,8 @@
 import treetensor.torch as ttorch
 from .base import choose_mark
 
+bool_init_dtype = torch.tensor([True, False]).dtype
+
 
 # noinspection DuplicatedCode,PyUnresolvedReferences
 class TestTorchTensorReduction:
@@ -14,7 +16,7 @@ def test_all(self):
             'b': {'x': [[True, True, ], [True, True, ]]}
         }).all()
         assert isinstance(t1, torch.Tensor)
-        assert t1.dtype == torch.bool
+        assert t1.dtype == bool_init_dtype
         assert t1
 
         t2 = ttorch.Tensor({
@@ -22,7 +24,7 @@ def test_all(self):
             'b': {'x': [[True, True, ], [True, True, ]]}
         }).all()
         assert isinstance(t2, torch.Tensor)
-        assert t2.dtype == torch.bool
+        assert t2.dtype == bool_init_dtype
         assert not t2
 
         t3 = ttorch.tensor({
@@ -48,7 +50,7 @@ def test_any(self):
             'b': {'x': [[False, False, ], [False, False, ]]}
         }).any()
         assert isinstance(t1, torch.Tensor)
-        assert t1.dtype == torch.bool
+        assert t1.dtype == bool_init_dtype
         assert t1
 
         t2 = ttorch.Tensor({
@@ -56,7 +58,7 @@ def test_any(self):
             'b': {'x': [[False, False, ], [False, False, ]]}
         }).any()
         assert isinstance(t2, torch.Tensor)
-        assert t2.dtype == torch.bool
+        assert t2.dtype == bool_init_dtype
         assert not t2
 
         t3 = ttorch.Tensor({
diff --git a/treetensor/common/module.py b/treetensor/common/module.py
index 3904f0d418..4c1ad226b6 100644
--- a/treetensor/common/module.py
+++ b/treetensor/common/module.py
@@ -30,7 +30,7 @@ def _load_func(name):
         @doc_from_base()
         @return_self_dec
         @post_process(auto_tree_cls)
-        @func_treelize(return_type=TreeValue, rise=True)
+        @func_treelize(return_type=TreeValue, subside=True, rise=True)
         @wraps(func, assigned=('__name__',), updated=())
         def _new_func(*args, **kwargs):
             return func(*args, **kwargs)
diff --git a/treetensor/common/wrappers.py b/treetensor/common/wrappers.py
index 6f0bfda6fc..297cf38d3a 100644
--- a/treetensor/common/wrappers.py
+++ b/treetensor/common/wrappers.py
@@ -1,7 +1,6 @@
 from functools import wraps
-from operator import itemgetter
 
-from treevalue import TreeValue, walk
+from treevalue import TreeValue, flatten_values
 
 __all__ = [
     'ireduce',
@@ -17,7 +16,7 @@ def _decorator(func):
         def _new_func(*args, **kwargs):
             result = func(*args, **kwargs)
             if isinstance(result, TreeValue):
-                it = map(itemgetter(1), walk(result, include_nodes=False))
+                it = flatten_values(result)
                 return rfunc(piter(it))
             else:
                 return result
diff --git a/treetensor/config/meta.py b/treetensor/config/meta.py
index 815d68c590..fe47d0f532 100644
--- a/treetensor/config/meta.py
+++ b/treetensor/config/meta.py
@@ -7,7 +7,7 @@
 __TITLE__ = "DI-treetensor"
 
 #: Version of this project.
-__VERSION__ = "0.2.0"
+__VERSION__ = "0.2.1"
 
 #: Short description of the project, will be included in ``setup.py``.
 __DESCRIPTION__ = 'A flexible, generalized tree-based tensor structure.'
diff --git a/treetensor/numpy/array.py b/treetensor/numpy/array.py
index 142c9587fc..470b61e4eb 100644
--- a/treetensor/numpy/array.py
+++ b/treetensor/numpy/array.py
@@ -1,4 +1,7 @@
+from functools import lru_cache
+
 import numpy
+import torch
 from treevalue import method_treelize
 
 from .base import TreeNumpy
@@ -12,6 +15,12 @@
 _ArrayProxy, _InstanceArrayProxy = get_tree_proxy(numpy.ndarray)
 
 
+@lru_cache()
+def _get_tensor_class(args0):
+    from ..torch import Tensor
+    return Tensor(args0)
+
+
 class _BaseArrayMeta(clsmeta(numpy.asarray, allow_dict=True)):
     pass
 
@@ -92,6 +101,13 @@ def all(self: numpy.ndarray, *args, **kwargs):
     def any(self: numpy.ndarray, *args, **kwargs):
         return self.any(*args, **kwargs)
 
+    @method_treelize(return_type=_get_tensor_class)
+    def tensor(self: numpy.ndarray, *args, **kwargs):
+        tensor_: torch.Tensor = torch.from_numpy(self)
+        if args or kwargs:
+            tensor_ = tensor_.to(*args, **kwargs)
+        return tensor_
+
     @method_treelize()
     def __eq__(self, other):
         """
diff --git a/treetensor/numpy/funcs.py b/treetensor/numpy/funcs.py
index b627bf8cfb..ba1d0ece2e 100644
--- a/treetensor/numpy/funcs.py
+++ b/treetensor/numpy/funcs.py
@@ -13,6 +13,8 @@
 __all__ = [
     'all', 'any', 'array',
     'equal', 'array_equal',
+    'stack', 'concatenate', 'split',
+    'zeros', 'ones',
 ]
 
 func_treelize = post_process(post_process(args_mapping(
@@ -71,3 +73,33 @@ def array(p_object, *args, **kwargs):
         })
     """
     return np.array(p_object, *args, **kwargs)
+
+
+@doc_from(np.stack)
+@func_treelize(subside=True)
+def stack(arrays, *args, **kwargs):
+    return np.stack(arrays, *args, **kwargs)
+
+
+@doc_from(np.concatenate)
+@func_treelize(subside=True)
+def concatenate(arrays, *args, **kwargs):
+    return np.concatenate(arrays, *args, **kwargs)
+
+
+@doc_from(np.split)
+@func_treelize(rise=True)
+def split(ary, *args, **kwargs):
+    return np.split(ary, *args, **kwargs)
+
+
+@doc_from(np.zeros)
+@func_treelize()
+def zeros(shape, *args, **kwargs):
+    return np.zeros(shape, *args, **kwargs)
+
+
+@doc_from(np.ones)
+@func_treelize()
+def ones(shape, *args, **kwargs):
+    return np.ones(shape, *args, **kwargs)