diff --git a/treetensor/common/__init__.py b/treetensor/common/__init__.py
index 6163b00278e6a7c385dc9415e67b7b8cc4f21350..04448872d945440151158e1a1c1058c4ab71a26c 100644
--- a/treetensor/common/__init__.py
+++ b/treetensor/common/__init__.py
@@ -1,3 +1,5 @@
+from .module import *
 from .object import *
+from .proxy import *
 from .trees import *
 from .wrappers import *
diff --git a/treetensor/common/module.py b/treetensor/common/module.py
new file mode 100644
index 0000000000000000000000000000000000000000..f840259879c5a891076b218a9ab798d6e465309c
--- /dev/null
+++ b/treetensor/common/module.py
@@ -0,0 +1,43 @@
+from functools import wraps
+from typing import Type
+
+from treevalue import TreeValue
+from treevalue import func_treelize as original_func_treelize
+from treevalue.tree.common import BaseTree
+from treevalue.utils import post_process
+
+from .trees import auto_tree
+from .wrappers import return_self
+from ..utils import doc_from_base as original_doc_from_base
+from ..utils import replaceable_partial, args_mapping
+
+__all__ = [
+    'module_func_loader',
+]
+
+
+def module_func_loader(base, cls: Type[TreeValue], module_name: str):
+    func_treelize = post_process(post_process(args_mapping(
+        lambda i, x: TreeValue(x) if isinstance(x, (dict, BaseTree, TreeValue)) else x)))(
+        replaceable_partial(original_func_treelize, return_type=cls)
+    )
+    doc_from_base = replaceable_partial(original_doc_from_base, base=base)
+    auto_tree_cls = replaceable_partial(auto_tree, cls=cls)
+
+    def _load_func(name):
+        func = getattr(base, name)
+        return_self_dec = return_self if func.__name__.endswith("_") else (lambda x: x)
+
+        @doc_from_base()
+        @return_self_dec
+        @post_process(auto_tree_cls)
+        @func_treelize(return_type=TreeValue, rise=True)
+        @wraps(func, assigned=('__name__',), updated=())
+        def _new_func(*args, **kwargs):
+            return func(*args, **kwargs)
+
+        _new_func.__qualname__ = _new_func.__name__
+        _new_func.__module__ = module_name
+        return _new_func
+
+    return _load_func
diff --git a/treetensor/common/proxy.py b/treetensor/common/proxy.py
new file mode 100644
index 0000000000000000000000000000000000000000..216688e85b2c85456821e74fb4b60562360ec755
--- /dev/null
+++ b/treetensor/common/proxy.py
@@ -0,0 +1,56 @@
+from functools import wraps
+from types import MethodType
+
+from treevalue import method_treelize, TreeValue
+from treevalue.utils import post_process
+
+from .trees import auto_tree
+from .wrappers import return_self
+from ..utils import doc_from_base as original_doc_from_base
+from ..utils import replaceable_partial
+
+__all__ = [
+    'get_tree_proxy',
+]
+
+
+def get_tree_proxy(base):
+    doc_from_base = replaceable_partial(original_doc_from_base, base=base)
+
+    class _TreeClassProxy:
+        def __init__(self, cls):
+            self.__torch_funcs = {}
+            self.__cls = cls
+
+        def __getattr__(self, name):
+            if name in self.__torch_funcs.keys():
+                return self.__torch_funcs[name]
+            elif hasattr(base, name) and not name.startswith('_') \
+                    and callable(getattr(base, name)):
+                _origin_func = getattr(base, name)
+                return_self_deco = return_self if name.endswith('_') else (lambda x: x)
+
+                @doc_from_base()
+                @return_self_deco
+                @post_process(lambda r: replaceable_partial(auto_tree, cls=self.__cls)(r))
+                @method_treelize(return_type=TreeValue, rise=True)
+                @wraps(_origin_func, assigned=('__name__',), updated=())
+                def _new_func(*args, **kwargs):
+                    return _origin_func(*args, **kwargs)
+
+                _new_func.__qualname__ = f'{self.__cls.__name__}.{name}'
+                _new_func.__module__ = self.__cls.__module__
+                self.__torch_funcs[name] = _new_func
+                return _new_func
+            else:
+                raise AttributeError(f'Function {repr(name)} not found in {repr(base)}')
+
+    class _TreeInstanceProxy:
+        def __init__(self, proxy, s):
+            self.__proxy = proxy
+            self.__self = s
+
+        def __getattr__(self, name):
+            return MethodType(getattr(self.__proxy, name), self.__self)
+
+    return _TreeClassProxy, _TreeInstanceProxy
diff --git a/treetensor/common/trees.py b/treetensor/common/trees.py
index 4a02ec3a79a5e6d7bac7188d497b831fd6e47009..b7def134439acfa49c6d9b8faab96edab310ad18 100644
--- a/treetensor/common/trees.py
+++ b/treetensor/common/trees.py
@@ -6,7 +6,7 @@ from typing import Optional, Tuple, Callable
 from typing import Type
 
 from treevalue import func_treelize as original_func_treelize
-from treevalue import general_tree_value, TreeValue
+from treevalue import general_tree_value, TreeValue, typetrans
 from treevalue.tree.common import BaseTree
 from treevalue.tree.tree.tree import get_data_property
 from treevalue.utils import post_process
@@ -15,7 +15,7 @@ from ..utils import replaceable_partial, args_mapping
 
 __all__ = [
     'BaseTreeStruct',
-    'print_tree', 'clsmeta',
+    'print_tree', 'clsmeta', 'auto_tree',
 ]
 
 
@@ -177,3 +177,18 @@ def clsmeta(func, allow_dict: bool = False) -> Type[type]:
                 return _result
 
     return _MetaClass
+
+
+# noinspection PyArgumentList
+def auto_tree(v, cls):
+    if isinstance(cls, type) and issubclass(cls, TreeValue):
+        cls = partial(typetrans, return_type=cls)
+
+    if isinstance(v, TreeValue):
+        return cls(v)
+    elif isinstance(v, (tuple, list, set)):
+        return type(v)((auto_tree(item, cls) for item in v))
+    elif isinstance(v, dict):
+        return type(v)({key: auto_tree(value, cls) for key, value in v.items()})
+    else:
+        return v
diff --git a/treetensor/numpy/__init__.py b/treetensor/numpy/__init__.py
index 46aa5d19273abba6041678241362e61c1f9cba4d..ddd0dbe373e1557a3a6c7be1c340f624b364f3a7 100644
--- a/treetensor/numpy/__init__.py
+++ b/treetensor/numpy/__init__.py
@@ -1,9 +1,56 @@
+import builtins
+from types import ModuleType, FunctionType, BuiltinFunctionType
+from typing import Iterable
+
+import numpy as np
+
 from .array import *
 from .array import __all__ as _array_all
 from .funcs import *
 from .funcs import __all__ as _funcs_all
+from .funcs import get_func_from_numpy
+from ..config.meta import __VERSION__
 
 __all__ = [
     *_funcs_all,
     *_array_all,
 ]
+
+_basic_types = (
+    builtins.bool, builtins.bytearray, builtins.bytes, builtins.complex, builtins.dict,
+    builtins.float, builtins.frozenset, builtins.int, builtins.list, builtins.range, builtins.set,
+    builtins.slice, builtins.str, builtins.tuple,
+)
+_np_all = set(np.__all__)
+
+
+class _Module(ModuleType):
+    def __init__(self, module):
+        ModuleType.__init__(self, module.__name__)
+
+        for name in filter(lambda x: x.startswith('__') and x.endswith('__'), dir(module)):
+            setattr(self, name, getattr(module, name))
+        self.__origin__ = module
+        self.__numpy_version__ = np.__version__
+        self.__version__ = __VERSION__
+
+    def __getattr__(self, name):
+        if (name in self.__all__) or \
+                (hasattr(self.__origin__, name) and isinstance(getattr(self.__origin__, name), ModuleType)):
+            return getattr(self.__origin__, name)
+        else:
+            item = getattr(np, name)
+            if isinstance(item, (FunctionType, BuiltinFunctionType)) and not name.startswith('_'):
+                return get_func_from_numpy(name)
+            elif isinstance(item, _basic_types) and name in _np_all:
+                return item
+            else:
+                raise AttributeError(f'Attribute {repr(name)} not found in {repr(__name__)}.')
+
+    def __dir__(self) -> Iterable[str]:
+        return self.__all__
+
+
+import sys
+
+sys.modules[__name__] = _Module(sys.modules[__name__])
diff --git a/treetensor/numpy/array.py b/treetensor/numpy/array.py
index 7cbcc9a8a1ba078aadb0d4b6af5ebee7b70fa87c..142c9587fc1a175db2670f1d702a179806db2bfe 100644
--- a/treetensor/numpy/array.py
+++ b/treetensor/numpy/array.py
@@ -1,52 +1,95 @@
-import numpy as np
+import numpy
 from treevalue import method_treelize
 
 from .base import TreeNumpy
-from ..common import Object, ireduce
+from ..common import Object, ireduce, clsmeta, get_tree_proxy
 from ..utils import current_names
 
 __all__ = [
     'ndarray'
 ]
 
+_ArrayProxy, _InstanceArrayProxy = get_tree_proxy(numpy.ndarray)
+
+
+class _BaseArrayMeta(clsmeta(numpy.asarray, allow_dict=True)):
+    pass
+
+
+# noinspection PyMethodParameters
+class _ArrayMeta(_BaseArrayMeta):
+    def __init__(cls, *args, **kwargs):
+        _BaseArrayMeta.__init__(cls, *args, **kwargs)
+        cls.__proxy = None
+
+    @property
+    def np(cls):
+        if not cls.__proxy:
+            cls.__proxy = _ArrayProxy(cls)
+        return cls.__proxy
+
+    def __getattr__(cls, name):
+        try:
+            return cls.np.__getattr__(name)
+        except AttributeError:
+            raise AttributeError(f"type object {repr(cls.__name__)} has no attribute {repr(name)}")
+
 
 # noinspection PyPep8Naming
 @current_names()
-class ndarray(TreeNumpy):
+class ndarray(TreeNumpy, metaclass=_ArrayMeta):
     """
     Overview:
         Real numpy tree.
     """
 
     @method_treelize(return_type=Object)
-    def tolist(self: np.ndarray):
+    def __get_attr(self, key):
+        return getattr(self, key)
+
+    def _attr_extern(self, name):
+        try:
+            return getattr(self.np, name)
+        except AttributeError:
+            tree = self.__get_attr(name)
+            if tree.map(lambda x: isinstance(x, numpy.ndarray)).all():
+                return tree.type(ndarray)
+            else:
+                return tree
+
+    @property
+    def np(self):
+        return _InstanceArrayProxy(self.__class__.np, self)
+
+    @method_treelize(return_type=Object)
+    def tolist(self: numpy.ndarray):
         return self.tolist()
 
     @property
     @ireduce(sum)
     @method_treelize(return_type=Object)
-    def size(self: np.ndarray) -> int:
+    def size(self: numpy.ndarray) -> int:
         return self.size
 
     @property
     @ireduce(sum)
     @method_treelize(return_type=Object)
-    def nbytes(self: np.ndarray) -> int:
+    def nbytes(self: numpy.ndarray) -> int:
         return self.nbytes
 
     @ireduce(sum)
     @method_treelize(return_type=Object)
-    def sum(self: np.ndarray, *args, **kwargs):
+    def sum(self: numpy.ndarray, *args, **kwargs):
         return self.sum(*args, **kwargs)
 
     @ireduce(all)
     @method_treelize(return_type=Object)
-    def all(self: np.ndarray, *args, **kwargs):
+    def all(self: numpy.ndarray, *args, **kwargs):
         return self.all(*args, **kwargs)
 
     @ireduce(any)
     @method_treelize(return_type=Object)
-    def any(self: np.ndarray, *args, **kwargs):
+    def any(self: numpy.ndarray, *args, **kwargs):
         return self.any(*args, **kwargs)
 
     @method_treelize()
diff --git a/treetensor/numpy/funcs.py b/treetensor/numpy/funcs.py
index c4f5a522b86dbe80b7b029d555d29225ef64bdc0..068b0371102d00546616b6b721174760da70f803 100644
--- a/treetensor/numpy/funcs.py
+++ b/treetensor/numpy/funcs.py
@@ -3,10 +3,11 @@ import builtins
 import numpy as np
 from treevalue import TreeValue
 from treevalue import func_treelize as original_func_treelize
+from treevalue.tree.common import BaseTree
 from treevalue.utils import post_process
 
 from .array import ndarray
-from ..common import ireduce, Object
+from ..common import ireduce, Object, module_func_loader
 from ..utils import replaceable_partial, doc_from, args_mapping
 
 __all__ = [
@@ -15,9 +16,10 @@ __all__ = [
 ]
 
 func_treelize = post_process(post_process(args_mapping(
-    lambda i, x: TreeValue(x) if isinstance(x, (dict, TreeValue)) else x)))(
+    lambda i, x: TreeValue(x) if isinstance(x, (dict, BaseTree, TreeValue)) else x)))(
     replaceable_partial(original_func_treelize, return_type=ndarray)
 )
+get_func_from_numpy = module_func_loader(np, ndarray, __name__)
 
 
 @doc_from(np.all)
diff --git a/treetensor/torch/base/torch.py b/treetensor/torch/base/torch.py
index c195b233d0db8313d8121fbf365f7ff9a8a45185..b2bbf2ca7b62bb68df61810eac6739748b4986fb 100644
--- a/treetensor/torch/base/torch.py
+++ b/treetensor/torch/base/torch.py
@@ -1,23 +1,9 @@
-from typing import Type
-
-from treevalue import TreeValue, typetrans
-
 from ...common import BaseTreeStruct
 
-__all__ = ['Torch', 'auto_torch']
+__all__ = [
+    'Torch'
+]
 
 
 class Torch(BaseTreeStruct):
     pass
-
-
-# noinspection PyArgumentList
-def auto_torch(v, cls: Type[Torch]):
-    if isinstance(v, TreeValue):
-        return typetrans(v, cls)
-    elif isinstance(v, (tuple, list, set)):
-        return type(v)((auto_torch(item, cls) for item in v))
-    elif isinstance(v, dict):
-        return type(v)({key: auto_torch(value, cls) for key, value in v.items()})
-    else:
-        return v
diff --git a/treetensor/torch/funcs/base.py b/treetensor/torch/funcs/base.py
index dae8e9d13b8933a7f2f4da377c7b90ea444730e5..1eab33875aa005c939e786972ba4f8e80ad0018b 100644
--- a/treetensor/torch/funcs/base.py
+++ b/treetensor/torch/funcs/base.py
@@ -1,14 +1,11 @@
-from functools import wraps
-
 import torch
 from treevalue import TreeValue
 from treevalue import func_treelize as original_func_treelize
 from treevalue.tree.common import BaseTree
 from treevalue.utils import post_process
 
-from ..base import auto_torch
 from ..tensor import Tensor
-from ...common import return_self
+from ...common import auto_tree, module_func_loader
 from ...utils import doc_from_base as original_doc_from_base
 from ...utils import replaceable_partial, args_mapping
 
@@ -17,23 +14,5 @@ func_treelize = post_process(post_process(args_mapping(
     replaceable_partial(original_func_treelize, return_type=Tensor)
 )
 doc_from_base = replaceable_partial(original_doc_from_base, base=torch)
-auto_tensor = replaceable_partial(auto_torch, cls=Tensor)
-
-_funcs_module = '.'.join(__name__.split('.')[:-1])
-
-
-def get_func_from_torch(name):
-    func = getattr(torch, name)
-    return_self_dec = return_self if func.__name__.endswith("_") else (lambda x: x)
-
-    @doc_from_base()
-    @return_self_dec
-    @post_process(auto_tensor)
-    @func_treelize(return_type=TreeValue, rise=True)
-    @wraps(func, assigned=('__name__',), updated=())
-    def _new_func(*args, **kwargs):
-        return func(*args, **kwargs)
-
-    _new_func.__qualname__ = _new_func.__name__
-    _new_func.__module__ = _funcs_module
-    return _new_func
+auto_tensor = replaceable_partial(auto_tree, cls=Tensor)
+get_func_from_torch = module_func_loader(torch, Tensor, '.'.join(__name__.split('.')[:-1]))
diff --git a/treetensor/torch/size.py b/treetensor/torch/size.py
index 2e86cfb9b54da89daa6a972beea4f0ef563b996a..31295744fa78c0a47ecbf010cab0ebdb30c62679 100644
--- a/treetensor/torch/size.py
+++ b/treetensor/torch/size.py
@@ -54,7 +54,7 @@ class Size(Torch, metaclass=clsmeta(torch.Size, allow_dict=True)):
         Examples::
 
             >>> import torch
-            >>> import treetensor.torch as ttorch
+            >>> import treetensor.numpy as ttorch
             >>> ttorch.Size([1, 2, 3])
             torch.Size([1, 2, 3])
 
@@ -81,7 +81,7 @@ class Size(Torch, metaclass=clsmeta(torch.Size, allow_dict=True)):
         Example::
 
             >>> import torch
-            >>> import treetensor.torch as ttorch
+            >>> import treetensor.numpy as ttorch
             >>> ttorch.Size({
             ...     'a': [1, 2],
             ...     'b': {'x': [3, 2, 4]},
@@ -99,7 +99,7 @@ class Size(Torch, metaclass=clsmeta(torch.Size, allow_dict=True)):
         Example::
 
             >>> import torch
-            >>> import treetensor.torch as ttorch
+            >>> import treetensor.numpy as ttorch
             >>> ttorch.Size({
             ...     'a': [1, 2],
             ...     'b': {'x': [3, 2, 4]},
@@ -132,7 +132,7 @@ class Size(Torch, metaclass=clsmeta(torch.Size, allow_dict=True)):
         Example::
 
             >>> import torch
-            >>> import treetensor.torch as ttorch
+            >>> import treetensor.numpy as ttorch
             >>> ttorch.Size({
             ...     'a': [1, 2],
             ...     'b': {'x': [3, 2, 4]},
diff --git a/treetensor/torch/tensor.py b/treetensor/torch/tensor.py
index 99754b36b7737d05b462a9f668f92871e5b48929..b6b970681da98b9ef5e9c38fd958422d7571305f 100644
--- a/treetensor/torch/tensor.py
+++ b/treetensor/torch/tensor.py
@@ -1,14 +1,11 @@
-from functools import wraps
-from types import MethodType
-
 import numpy as np
 import torch as pytorch
 from treevalue import method_treelize, TreeValue
 from treevalue.utils import post_process
 
-from .base import Torch, auto_torch, rmreduce, post_reduce, auto_reduce
+from .base import Torch, rmreduce, post_reduce, auto_reduce
 from .size import Size
-from ..common import Object, ireduce, clsmeta, return_self
+from ..common import Object, ireduce, clsmeta, return_self, auto_tree, get_tree_proxy
 from ..numpy import ndarray
 from ..utils import current_names, class_autoremove, replaceable_partial
 from ..utils import doc_from_base as original_doc_from_base
@@ -18,6 +15,7 @@ __all__ = [
 ]
 
 doc_from_base = replaceable_partial(original_doc_from_base, base=pytorch.Tensor)
+_TorchProxy, _InstanceTorchProxy = get_tree_proxy(pytorch.Tensor)
 
 
 def _to_tensor(*args, **kwargs):
@@ -30,44 +28,6 @@ def _to_tensor(*args, **kwargs):
     return pytorch.tensor(*args, **kwargs)
 
 
-class _TorchProxy:
-    def __init__(self, cls):
-        self.__torch_funcs = {}
-        self.__cls = cls
-
-    def __getattr__(self, name):
-        if name in self.__torch_funcs.keys():
-            return self.__torch_funcs[name]
-        elif hasattr(pytorch.Tensor, name) and not name.startswith('_') \
-                and callable(getattr(pytorch.Tensor, name)):
-            _origin_func = getattr(pytorch.Tensor, name)
-            return_self_deco = return_self if name.endswith('_') else (lambda x: x)
-
-            @doc_from_base()
-            @return_self_deco
-            @post_process(lambda r: replaceable_partial(auto_torch, cls=self.__cls)(r))
-            @method_treelize(return_type=TreeValue, rise=True)
-            @wraps(_origin_func, assigned=('__name__',), updated=())
-            def _new_func(*args, **kwargs):
-                return _origin_func(*args, **kwargs)
-
-            _new_func.__qualname__ = f'{self.__cls.__name__}.{name}'
-            _new_func.__module__ = self.__cls.__module__
-            self.__torch_funcs[name] = _new_func
-            return _new_func
-        else:
-            raise AttributeError(f'Function {repr(name)} not found in {repr(pytorch)}')
-
-
-class _InstanceTorchProxy:
-    def __init__(self, proxy, s):
-        self.__proxy = proxy
-        self.__self = s
-
-    def __getattr__(self, name):
-        return MethodType(getattr(self.__proxy, name), self.__self)
-
-
 class _BaseTensorMeta(clsmeta(_to_tensor, allow_dict=True)):
     pass
 
@@ -76,13 +36,13 @@ class _BaseTensorMeta(clsmeta(_to_tensor, allow_dict=True)):
 class _TensorMeta(_BaseTensorMeta):
     def __init__(cls, *args, **kwargs):
         _BaseTensorMeta.__init__(cls, *args, **kwargs)
-        cls.__torch_proxy = None
+        cls.__proxy = None
 
     @property
     def torch(cls):
-        if not cls.__torch_proxy:
-            cls.__torch_proxy = _TorchProxy(cls)
-        return cls.__torch_proxy
+        if not cls.__proxy:
+            cls.__proxy = _TorchProxy(cls)
+        return cls.__proxy
 
     def __getattr__(cls, name):
         try:
@@ -439,7 +399,7 @@ class Tensor(Torch, metaclass=_TensorMeta):
         return self
 
     # noinspection PyShadowingBuiltins
-    @post_process(lambda r: replaceable_partial(auto_torch, cls=Tensor)(r))
+    @post_process(lambda r: replaceable_partial(auto_tree, cls=Tensor)(r))
     @method_treelize(return_type=TreeValue, rise=True)
     def __max_nr(self, *args, **kwargs):
         return pytorch.max(self, *args, **kwargs)
@@ -459,7 +419,7 @@ class Tensor(Torch, metaclass=_TensorMeta):
         return self
 
     # noinspection PyShadowingBuiltins
-    @post_process(lambda r: replaceable_partial(auto_torch, cls=Tensor)(r))
+    @post_process(lambda r: replaceable_partial(auto_tree, cls=Tensor)(r))
     @method_treelize(return_type=TreeValue, rise=True)
     def __min_nr(self, *args, **kwargs):
         return pytorch.min(self, *args, **kwargs)
@@ -479,7 +439,7 @@ class Tensor(Torch, metaclass=_TensorMeta):
         return self
 
     # noinspection PyShadowingBuiltins
-    @post_process(lambda r: replaceable_partial(auto_torch, cls=Tensor)(r))
+    @post_process(lambda r: replaceable_partial(auto_tree, cls=Tensor)(r))
     @method_treelize(return_type=TreeValue, rise=True)
     def __sum_nr(self, *args, **kwargs):
         return pytorch.sum(self, *args, **kwargs)
@@ -922,7 +882,7 @@ class Tensor(Torch, metaclass=_TensorMeta):
         return self.log10_(*args, **kwargs)
 
     @doc_from_base()
-    @post_process(lambda r: replaceable_partial(auto_torch, cls=Tensor)(r))
+    @post_process(lambda r: replaceable_partial(auto_tree, cls=Tensor)(r))
     @method_treelize(return_type=TreeValue, rise=True)
     def split(self, split_size, *args, **kwargs):
         """
@@ -931,7 +891,7 @@ class Tensor(Torch, metaclass=_TensorMeta):
         return self.split(split_size, *args, **kwargs)
 
     @doc_from_base()
-    @post_process(lambda r: replaceable_partial(auto_torch, cls=Tensor)(r))
+    @post_process(lambda r: replaceable_partial(auto_tree, cls=Tensor)(r))
     @method_treelize(return_type=TreeValue, rise=True)
     def chunk(self, chunks, *args, **kwargs):
         """