ёiSSKJr SSKJr SSKJr SSKrSSKJr SSKJ r J r J r J r \(aSSK Jr SSKJr SS KJr SS jr\ SS j5r\ SSS jj5r\ SSS jj5rg)) annotations)reduce) TYPE_CHECKINGN)_C_ops)_create_tensor_dygraph_tracer dygraph_onlyin_dygraph_mode)Iterable)Tensor) ShapeLikecJ[SS9n[5(aY[RRR 5 [ R"X5nURU5 SSS5 g[5RSSU0XS.SU0SS9 g!,(df  g=f) Nint64dtypereshape2X)OutXShapeshapeTtypeinputsoutputsattrs stop_gradient) rr paddlebasedygraphno_gradrreshape_share_underline_tensor_tortrace_op)xrx_shapetmp_outs d/var/www/html/banglarbhumi/venv/lib/python3.13/site-packages/paddle/nn/utils/transform_parameters.py_inplace_reshape_dygraphr(%s7+G [[ ( ( *nnQ.G  . .q 1+ * ""81E" # + *s (B B"cz[UR5S:XdeURSURS/n[RRR 5 [R "X5n[R"USS/5nURU5 SSS5 g!,(df  g=f)a A tool function. Permute date of parameter as a 'columns' stride. Now, it only support 2-D parameter. Args: param(Tensor): The param that will be strided according to 'columns'. Examples: .. code-block:: python >>> import paddle >>> paddle.seed(100) >>> linear = paddle.nn.Linear(2, 3) >>> print(linear.weight) Parameter containing: Tensor(shape=[2, 3], dtype=float32, place=Place(cpu), stop_gradient=False, [[ 0.11732829, -0.64161885, -1.06996548], [ 0.03456247, -0.29862350, -0.52380574]]) >>> paddle.nn.utils._stride_column(linear.weight) >>> print(linear.weight) rN) lenrrrrr r! transposer")paramr reshape_var transpose_vars r'_stride_columnr15s2 u{{ q  [[^U[[^ ,E    $ $ &nnU2 ((q!f= 007 ' & &s "AB,, B:c&USRn/nUH+nURUR5 [US/5 M- [ US9n[ 5(aZ[ RRR5 [R"US5nURU5 SSS5 O"[5RSSU0SU/0SS0S S 9 [U5Hupt[XCU5 M S UlU$!,(df  N9=f) a Flatten parameters to a 1-D Tensor. Args: parameters(Iterable[Tensor]): Iterable Tensors that are trainable parameters of a Layer. name(str, optional): The default value is None. Normally there is no need for user to set this property. For more information, please refer to :ref:`api_guide_Name`. Returns: A 1-D Tensor, which represents the parameters of a Layer. Examples: .. code-block:: pycon >>> import paddle >>> paddle.seed(2023) >>> linear = paddle.nn.Linear(10, 15) >>> t = paddle.nn.utils.parameters_to_vector(linear.parameters()) >>> print(t.shape) paddle.Size([165]) rrNconcatrraxisTrF)rappendrr(rr rrrr rr4r"rr# enumerater) parametersnamer origin_shapesr.outtmpis r'parameters_to_vectorr>Vs8 qM  EMU[[) - u %C [[ ( ( *-- A.C  * *3 /+ * ""$SEN1+ # j) a(89*C J+ *s )D Dc[UR5S:Xde/n/nSnUHCnURnURU5 [SUS5nXX- nURU5 ME [U5S:XaURS5 [ 5(a[ R RR5 /n XPRS:Xa[R"XS5n OBXPRS:a%Sn UHn U RX X-5 X- n M O [S5e[S[U55Hn XRX5 M SSS5 O#[5RSSU/0SU0SUS .S S 9 [!U5Hup[#XcU 5 M g!,(df  N1=f) a Transform a 1-D Tensor to the input ``parameters`` . Args: vec (Tensor): A 1-D Tensor, which will be sliced and copied to the input ``parameters`` . parameters (Iterable[Tensor]): Iterable Tensors that are trainable parameters of a Layer. name(str, optional): The default value is None. Normally there is no need for user to set this property. For more information, please refer to :ref:`api_guide_Name`. Examples: .. code-block:: python >>> import paddle >>> weight_attr = paddle.ParamAttr(initializer=paddle.nn.initializer.Constant(3.)) >>> linear1 = paddle.nn.Linear(10, 15, weight_attr) >>> vec = paddle.nn.utils.parameters_to_vector(linear1.parameters()) >>> linear2 = paddle.nn.Linear(10, 15) >>> # copy weight of linear1 to linear2 >>> paddle.nn.utils.vector_to_parameters(vec, linear2.parameters()) >>> print((linear1.weight == linear2.weight).all()) Tensor(shape=[], dtype=bool, place=Place(cpu), stop_gradient=True, True) r+rc X-$N)r$ys r'&vector_to_parameters..sAEzaThe total_elements of vec should be equal to or larger than the number of elements in parameters.Nsplitrr)r5sectionsTr)r,rr6rr rrrr rrG ValueErrorranger"rr#r7r() vecr8r9r:rHtotal_elementsr.rnumelrespointersectionr=s r'vector_to_parametersrQs: syy>Q  MHN U#)5!4  8} [[ ( ( *C1-ll3!4))A,.'GJJsW->?@&G (!w1c*o.11*-@/+ * ""#<J'(3 # j) a(89*1+ *s B F11 F?)r$r rr returnNone)r.r rRrSrA)r8Iterable[Tensor]r9 str | NonerRr )rKr r8rTr9rUrRrS) __future__r functoolsrtypingrrrpaddle.base.frameworkrrr r collections.abcr r paddle._typingr r(r1r>rQrBrFr'r\s# ((  88@591 1(21 11hBFD: D:-D:5?D: D:D:rF