maxframe 2.4.0rc1__cp312-cp312-win32.whl

maxframe/tensor/datasource/diag.py

@@ -0,0 +1,140 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import List
+
+from ... import opcodes
+from ...core import EntityData
+from ...lib.sparse import diag as sparse_diag
+from ...lib.sparse import get_sparse_module, issparse
+from ...lib.sparse.core import get_array_module
+from ...serialization.serializables import Int32Field
+from ..core import TENSOR_TYPE, TensorOrder
+from .array import tensor
+from .core import TensorHasInput
+
+
+def _get_diag_shape(v_shape, k):
+    size_0, size_1 = 0, 0
+    if k > 0:
+        size_1 += k
+    elif k < 0:
+        size_0 -= k
+    size = min(v_shape[0] - size_0, v_shape[1] - size_1)
+    return (size,)
+
+
+class TensorDiag(TensorHasInput):
+    _op_type_ = opcodes.TENSOR_DIAG
+
+    k = Int32Field("k", default=None)
+
+    @classmethod
+    def _set_inputs(cls, op: "TensorDiag", inputs: List[EntityData]):
+        super()._set_inputs(op, inputs)
+        if op.dtype is None:
+            op._dtype = op.inputs[0].dtype
+
+    def __call__(self, v, shape, chunk_size=None):
+        return self.new_tensor(
+            [v], shape, raw_chunk_size=chunk_size, order=TensorOrder.C_ORDER
+        )
+
+
+def diag(v, k=0, sparse=None, gpu=None, chunk_size=None):
+    """
+    Extract a diagonal or construct a diagonal tensor.
+
+    See the more detailed documentation for ``mt.diagonal`` if you use this
+    function to extract a diagonal and wish to write to the resulting tensor
+
+    Parameters
+    ----------
+    v : array_like
+        If `v` is a 2-D tensor, return its `k`-th diagonal.
+        If `v` is a 1-D tensor, return a 2-D tensor with `v` on the `k`-th
+        diagonal.
+    k : int, optional
+        Diagonal in question. The default is 0. Use `k>0` for diagonals
+        above the main diagonal, and `k<0` for diagonals below the main
+        diagonal.
+    sparse: bool, optional
+        Create sparse tensor if True, False as default
+    gpu : bool, optional
+        Allocate the tensor on GPU if True, False as default
+    chunk_size : int or tuple of int or tuple of ints, optional
+        Desired chunk size on each dimension
+
+    Returns
+    -------
+    out : Tensor
+        The extracted diagonal or constructed diagonal tensor.
+
+    See Also
+    --------
+    diagonal : Return specified diagonals.
+    diagflat : Create a 2-D array with the flattened input as a diagonal.
+    trace : Sum along diagonals.
+    triu : Upper triangle of a tensor.
+    tril : Lower triangle of a tensor.
+
+    Examples
+    --------
+    >>> import maxframe.tensor as mt
+
+    >>> x = mt.arange(9).reshape((3,3))
+    >>> x.execute()
+    array([[0, 1, 2],
+           [3, 4, 5],
+           [6, 7, 8]])
+
+    >>> mt.diag(x).execute()
+    array([0, 4, 8])
+    >>> mt.diag(x, k=1).execute()
+    array([1, 5])
+    >>> mt.diag(x, k=-1).execute()
+    array([3, 7])
+
+    >>> mt.diag(mt.diag(x)).execute()
+    array([[0, 0, 0],
+           [0, 4, 0],
+           [0, 0, 8]])
+
+    """
+    if not isinstance(v, TENSOR_TYPE):
+        tensor_v = tensor(v)
+        if tensor_v.issparse():
+            xps = get_sparse_module(tensor_v.data)
+            v = xps.csr_matrix(
+                (tensor_v.op.data, tensor_v.op.indices, tensor_v.op.indptr),
+                tensor_v.shape,
+            )
+            diag_v = sparse_diag(v, k=k)
+        else:
+            v = tensor(v).op.data
+            diag_v = get_array_module(v).diag(v, k=k)
+        sparse = sparse if sparse is not None else issparse(v)
+        return tensor(diag_v, gpu=gpu, sparse=sparse, chunk_size=chunk_size)
+
+    sparse = sparse if sparse is not None else v.issparse()
+
+    if v.ndim == 1:
+        shape = (v.size + abs(k),) * 2
+    elif v.ndim == 2:
+        shape = _get_diag_shape(v.shape, k)
+    else:
+        raise ValueError("Input must be 1- or 2-d.")
+
+    op = TensorDiag(k=k, dtype=v.dtype, gpu=gpu, sparse=sparse)
+    return op(v, shape)

maxframe/tensor/datasource/diagflat.py

@@ -0,0 +1,69 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from ..core import Tensor
+from .array import tensor as astensor
+from .diag import diag
+
+
+def diagflat(v, k=0, sparse=None, gpu=None, chunk_size=None):
+    """
+    Create a two-dimensional tensor with the flattened input as a diagonal.
+
+    Parameters
+    ----------
+    v : array_like
+        Input data, which is flattened and set as the `k`-th
+        diagonal of the output.
+    k : int, optional
+        Diagonal to set; 0, the default, corresponds to the "main" diagonal,
+        a positive (negative) `k` giving the number of the diagonal above
+        (below) the main.
+    sparse: bool, optional
+        Create sparse tensor if True, False as default
+    gpu : bool, optional
+        Allocate the tensor on GPU if True, False as default
+    chunk_size : int or tuple of int or tuple of ints, optional
+        Desired chunk size on each dimension
+
+    Returns
+    -------
+    out : Tensor
+        The 2-D output tensor.
+
+    See Also
+    --------
+    diag : MATLAB work-alike for 1-D and 2-D tensors.
+    diagonal : Return specified diagonals.
+    trace : Sum along diagonals.
+
+    Examples
+    --------
+    >>> import maxframe.tensor as mt
+
+    >>> mt.diagflat([[1,2], [3,4]]).execute()
+    array([[1, 0, 0, 0],
+           [0, 2, 0, 0],
+           [0, 0, 3, 0],
+           [0, 0, 0, 4]])
+
+    >>> mt.diagflat([1,2], 1).execute()
+    array([[0, 1, 0],
+           [0, 0, 2],
+           [0, 0, 0]])
+
+    """
+    if not isinstance(v, Tensor):
+        v = astensor(v).op.data
+    return diag(v.flatten(), k=k, sparse=sparse, gpu=gpu, chunk_size=chunk_size)

maxframe/tensor/datasource/empty.py

@@ -0,0 +1,167 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import numpy as np
+
+from ... import opcodes
+from ...serialization.serializables import KeyField, StringField
+from ..utils import get_order
+from .array import tensor
+from .core import TensorLike, TensorNoInput
+
+
+class TensorEmptyBase(object):
+    __slots__ = ()
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._gen_rand()
+
+    def _gen_rand(self):
+        if getattr(self, "_rand", None) is None:
+            self._obj_set("_rand", np.random.random())
+
+
+class TensorEmpty(TensorEmptyBase, TensorNoInput):
+    __slots__ = ("_rand",)
+    _op_type_ = opcodes.TENSOR_EMPTY
+
+    order = StringField("order", default=None)
+
+    def __init__(self, dtype=None, **kw):
+        dtype = np.dtype(dtype or "f8")
+        super().__init__(dtype=dtype, **kw)
+
+
+def empty(shape, dtype=None, chunk_size=None, gpu=None, order="C"):
+    """
+    Return a new tensor of given shape and type, without initializing entries.
+
+    Parameters
+    ----------
+    shape : int or tuple of int
+        Shape of the empty tensor
+    dtype : data-type, optional
+        Desired output data-type.
+    chunk_size : int or tuple of int or tuple of ints, optional
+        Desired chunk size on each dimension
+    gpu : bool, optional
+        Allocate the tensor on GPU if True, False as default
+    order : {'C', 'F'}, optional, default: 'C'
+        Whether to store multi-dimensional data in row-major
+        (C-style) or column-major (Fortran-style) order in
+        memory.
+
+    Returns
+    -------
+    out : Tensor
+        Tensor of uninitialized (arbitrary) data of the given shape, dtype, and
+        order.  Object arrays will be initialized to None.
+
+    See Also
+    --------
+    empty_like, zeros, ones
+
+    Notes
+    -----
+    `empty`, unlike `zeros`, does not set the array values to zero,
+    and may therefore be marginally faster.  On the other hand, it requires
+    the user to manually set all the values in the array, and should be
+    used with caution.
+
+    Examples
+    --------
+    >>> import maxframe.tensor as mt
+    >>> mt.empty([2, 2]).execute()
+    array([[ -9.74499359e+001,   6.69583040e-309],
+           [  2.13182611e-314,   3.06959433e-309]])         #random
+    >>> mt.empty([2, 2], dtype=int).execute()
+    array([[-1073741821, -1067949133],
+           [  496041986,    19249760]])                     #random
+    """
+    tensor_order = get_order(
+        order,
+        None,
+        available_options="CF",
+        err_msg="only 'C' or 'F' order is permitted",
+    )
+    op = TensorEmpty(dtype=dtype, gpu=gpu, order=order)
+    return op(shape, chunk_size=chunk_size, order=tensor_order)
+
+
+class TensorEmptyLike(TensorEmptyBase, TensorLike):
+    __slots__ = ("_rand",)
+    _op_type_ = opcodes.TENSOR_EMPTY_LIKE
+
+    _input = KeyField("input")
+    order = StringField("order", default=None)
+
+    def __init__(self, dtype=None, gpu=None, sparse=False, **kw):
+        dtype = np.dtype(dtype) if dtype is not None else None
+        super().__init__(dtype=dtype, gpu=gpu, sparse=sparse, **kw)
+
+
+def empty_like(a, dtype=None, gpu=None, order="K"):
+    """
+    Return a new tensor with the same shape and type as a given tensor.
+
+    Parameters
+    ----------
+    a : array_like
+        The shape and data-type of `a` define these same attributes of the
+        returned tensor.
+    dtype : data-type, optional
+        Overrides the data type of the result.
+    gpu : bool, optional
+        Allocate the tensor on GPU if True, None as default
+    order : {'C', 'F', 'A', or 'K'}, optional
+        Overrides the memory layout of the result. 'C' means C-order,
+        'F' means F-order, 'A' means 'F' if ``prototype`` is Fortran
+        contiguous, 'C' otherwise. 'K' means match the layout of ``prototype``
+        as closely as possible.
+
+    Returns
+    -------
+    out : Tensor
+        Array of uninitialized (arbitrary) data with the same
+        shape and type as `a`.
+    See Also
+    --------
+    ones_like : Return a tensor of ones with shape and type of input.
+    zeros_like : Return a tensor of zeros with shape and type of input.
+    empty : Return a new uninitialized tensor.
+    ones : Return a new tensor setting values to one.
+    zeros : Return a new tensor setting values to zero.
+    Notes
+    -----
+    This function does *not* initialize the returned tensor; to do that use
+    `zeros_like` or `ones_like` instead.  It may be marginally faster than
+    the functions that do set the array values.
+    Examples
+    --------
+    >>> import maxframe.tensor as mt
+    >>> a = ([1,2,3], [4,5,6])                         # a is array-like
+    >>> mt.empty_like(a).execute()
+    array([[-1073741821, -1073741821,           3],    #ranm
+           [          0,           0, -1073741821]])
+    >>> a = mt.array([[1., 2., 3.],[4.,5.,6.]])
+    >>> mt.empty_like(a).execute()
+    array([[ -2.00000715e+000,   1.48219694e-323,  -2.00000572e+000],#random
+           [  4.38791518e-305,  -2.00000715e+000,   4.17269252e-309]])
+    """
+    a = tensor(a)
+    tensor_order = get_order(order, a.order)
+    gpu = a.op.gpu if gpu is None else gpu
+    op = TensorEmptyLike(dtype=dtype, gpu=gpu, sparse=a.issparse(), order=order)
+    return op(a, order=tensor_order)

maxframe/tensor/datasource/eye.py

@@ -0,0 +1,95 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import numpy as np
+
+from ... import opcodes
+from ...serialization.serializables import Int32Field, StringField
+from ..utils import get_order
+from .core import TensorNoInput
+
+
+class TensorEye(TensorNoInput):
+    _op_type_ = opcodes.TENSOR_EYE
+
+    k = Int32Field("k", default=0)
+    order = StringField("order", default=None)
+
+    def __init__(self, dtype=None, **kw):
+        dtype = np.dtype(dtype or "f8")
+        super().__init__(dtype=dtype, **kw)
+
+
+def eye(N, M=None, k=0, dtype=None, sparse=False, gpu=None, chunk_size=None, order="C"):
+    """
+    Return a 2-D tensor with ones on the diagonal and zeros elsewhere.
+
+    Parameters
+    ----------
+    N : int
+      Number of rows in the output.
+    M : int, optional
+      Number of columns in the output. If None, defaults to `N`.
+    k : int, optional
+      Index of the diagonal: 0 (the default) refers to the main diagonal,
+      a positive value refers to an upper diagonal, and a negative value
+      to a lower diagonal.
+    dtype : data-type, optional
+      Data-type of the returned tensor.
+    sparse: bool, optional
+        Create sparse tensor if True, False as default
+    gpu : bool, optional
+        Allocate the tensor on GPU if True, False as default
+    chunk_size : int or tuple of int or tuple of ints, optional
+        Desired chunk size on each dimension
+    order : {'C', 'F'}, optional
+        Whether the output should be stored in row-major (C-style) or
+        column-major (Fortran-style) order in memory.
+
+    Returns
+    -------
+    I : Tensor of shape (N,M)
+      An tensor where all elements are equal to zero, except for the `k`-th
+      diagonal, whose values are equal to one.
+
+    See Also
+    --------
+    identity : (almost) equivalent function
+    diag : diagonal 2-D tensor from a 1-D tensor specified by the user.
+
+    Examples
+    --------
+    >>> import maxframe.tensor as mt
+
+    >>> mt.eye(2, dtype=int).execute()
+    array([[1, 0],
+           [0, 1]])
+    >>> mt.eye(3, k=1).execute()
+    array([[ 0.,  1.,  0.],
+           [ 0.,  0.,  1.],
+           [ 0.,  0.,  0.]])
+
+    """
+    if M is None:
+        M = N
+
+    shape = (N, M)
+    tensor_order = get_order(
+        order,
+        None,
+        available_options="CF",
+        err_msg="only 'C' or 'F' order is permitted",
+    )
+    op = TensorEye(k=k, dtype=dtype, gpu=gpu, sparse=sparse, order=order)
+    return op(shape, chunk_size=chunk_size, order=tensor_order)

maxframe/tensor/datasource/from_dataframe.py

@@ -0,0 +1,68 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from ... import opcodes
+from ...serialization.serializables import BoolField, KeyField
+from ..core import TensorOrder
+from ..utils import to_numpy
+from .core import TensorHasInput
+
+
+class TensorFromDataFrame(TensorHasInput):
+    """represent tensor from DataFrame"""
+
+    _op_type_ = opcodes.TENSOR_FROM_DATAFRAME
+    _input = KeyField("_input")
+    extract_multi_index = BoolField("extract_multi_index", default=None)
+
+    def __call__(self, a, order=None):
+        from ...dataframe.core import INDEX_TYPE, IndexValue
+
+        if (
+            self.extract_multi_index
+            and isinstance(a, INDEX_TYPE)
+            and isinstance(a.index_value.value, IndexValue.MultiIndex)
+        ):
+            order = a.order if order is None else order
+            return self.new_tensor(
+                [a], (a.shape[0], len(a.index_value.value.names)), order=order
+            )
+        else:
+            self.extract_multi_index = False
+
+        return super().__call__(a, order=order)
+
+
+def from_dataframe(in_df, dtype=None):
+    from ...dataframe.utils import build_empty_df
+
+    if dtype is None:
+        empty_pdf = build_empty_df(in_df.dtypes)
+        dtype = to_numpy(empty_pdf).dtype
+    op = TensorFromDataFrame(dtype=dtype, gpu=in_df.op.gpu)
+    return op(in_df, order=TensorOrder.F_ORDER)  # return tensor with F-order always
+
+
+def from_series(in_series, dtype=None):
+    op = TensorFromDataFrame(dtype=dtype or in_series.dtype, gpu=in_series.op.gpu)
+    return op(in_series, order=TensorOrder.F_ORDER)  # return tensor with F-order always
+
+
+def from_index(in_index, dtype=None, extract_multi_index=False):
+    op = TensorFromDataFrame(
+        dtype=dtype or in_index.dtype,
+        gpu=in_index.op.gpu,
+        extract_multi_index=extract_multi_index,
+    )
+    return op(in_index, order=TensorOrder.F_ORDER)  # return tensor with F-order always

maxframe/tensor/datasource/from_dense.py

@@ -0,0 +1,37 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from ... import opcodes
+from ...serialization.serializables import AnyField, KeyField
+from .array import tensor
+from .core import TensorHasInput
+
+
+class DenseToSparse(TensorHasInput):
+    _op_type_ = opcodes.DENSE_TO_SPARSE
+
+    _input = KeyField("input")
+    missing = AnyField("missing", default=None)
+
+    def __init__(self, **kw):
+        super().__init__(sparse=True, **kw)
+
+
+def fromdense(a, missing=None):
+    a = tensor(a)
+    if a.issparse():
+        return a
+
+    op = DenseToSparse(dtype=a.dtype, gpu=a.op.gpu, missing=missing)
+    return op(a)

maxframe/tensor/datasource/from_sparse.py

@@ -0,0 +1,45 @@
+# Copyright 1999-2025 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from ... import opcodes
+from ...serialization.serializables import AnyField, KeyField, StringField
+from ..utils import get_order
+from .array import tensor
+from .core import TensorHasInput
+
+
+class SparseToDense(TensorHasInput):
+    _op_type_ = opcodes.SPARSE_TO_DENSE
+
+    _input = KeyField("input")
+    order = StringField("order", default=None)
+    fill_value = AnyField("fill_value", default=None)
+
+    def __init__(self, **kw):
+        super().__init__(sparse=False, **kw)
+
+
+def fromsparse(a, order="C", fill_value=None):
+    a = tensor(a)
+    if not a.issparse():
+        return a.astype(a.dtype, order=order, copy=False)
+
+    tensor_order = get_order(
+        order,
+        None,
+        available_options="CF",
+        err_msg="only 'C' or 'F' order is permitted",
+    )
+    op = SparseToDense(dtype=a.dtype, gpu=a.op.gpu, order=order, fill_value=fill_value)
+    return op(a, order=tensor_order)