[ThXRSSKrSSKrSSKJr SSKJrJrJrJr SSK r SSK J s J r SSK JrJr SSKJr SSKJrJr SSKJrJr SSKJrJrJrJr /S QrS rS r SZS \S \\!\"\!4S\!S\#\S44Sjjr$S\S\4Sjr%\(a SS.S\\\"\4S\\&S\#\S44Sjjr'OSS.S\\&S\#\S44Sjjr'S\\&4Sjr(S[S\S\!S\\!S\\!S\\S \)S!\&S"\)S#\\)S$\\)S%\\)S\4S&jjr*\"\ RVS'5r+\(a\r,O\#\\\4r,S\S\S(\)S)\)S*\)S\\!S\,4 S+jjr-S]S\S)\)S*\)S\\!S\,4 S,jjr.S\S-jr/S\S.jr0S\S/jr1\"S*S0S\/\0\2S1S29r3\"S*S0S\-\1\2S1S29r4\"S)S3S\4\3\2S1S29r5\-Rl\5l6S]S4jr7S]S5jr8S]S6jr9\"S*S7S\7\8\2S8S29r:\"S*S7S\.\9\2S8S29r;\"S)S3S\;\:\2S8S29r<\.Rl\jj5r=S^S:\\ R4S?jjr=S\S\4S@jr>SAr?S`SBjr@SCrASDrBSErC\(aO<\SaSFj5rD\SaSGj5rD\SaSHj5rD\SaSIj5rDSaSJ\\\E\&44SKjjrDSL\SM\\!\\!\ R4S\#\S44SNjrGSL\SM\\!\\!4S\4SOjrHSSP.SQjrISbSRjrJ\(a\\rKO\"\rKSS\!ST\)S:\KSS4SUjrLSbSVjrMSbSWjrN\"STS3S\M\N\2SXS29rO\JRl\Ol6SYrPg)cN)Sequence)AnyOptional TYPE_CHECKINGUnion)_VFTensor) _add_docstr) _overloadboolean_dispatch) pca_lowrank svd_lowrank)handle_torch_functionhas_torch_functionhas_torch_function_unaryhas_torch_function_variadic) atleast_1d atleast_2d atleast_3d align_tensorsbroadcast_shapesbroadcast_tensorscartesian_prod block_diagcdist chain_matmuleinsumistftlunormmeshgridr splitstftr tensordotuniqueunique_consecutive unravel_indexcp[U5(a[[U/UQ76$[R"U5$)abroadcast_tensors(*tensors) -> List of Tensors Broadcasts the given tensors according to :ref:`broadcasting-semantics`. Args: *tensors: any number of tensors of the same type .. warning:: More than one element of a broadcasted tensor may refer to a single memory location. As a result, in-place operations (especially ones that are vectorized) may result in incorrect behavior. If you need to write to the tensors, please clone them first. Example:: >>> x = torch.arange(3).view(1, 3) >>> y = torch.arange(2).view(2, 1) >>> a, b = torch.broadcast_tensors(x, y) >>> a.size() torch.Size([2, 3]) >>> a tensor([[0, 1, 2], [0, 1, 2]]) )rrrrtensorss H/var/www/auris/envauris/lib/python3.13/site-packages/torch/functional.pyrr0s36'""$%6J'JJ   ))c [RR5(GdSnUHen[U[[R 45(a US:aSnM2M4[U[ [45(dMQ[U5nX:dMcUnMg S/U-nSSK J nJ n UGHn[U[[R 45(aU4n[U[ [45(a[SS[U5- S5HnX'S:a[SX'SX'S35eU"X'5(a&U"XG5(aU"X'XG:H5(aMUO(U"X'S:H5(dU"X'XG:H5(aM~XGS:wa [S5eX'XG'M M[S U5e [R"U5$[R"5 [R "S S S 9nUVs/sHo(R#U5PM n n[%U 6n U SR&sS S S 5 $s snf!,(df  g =f)abroadcast_shapes(*shapes) -> Size Similar to :func:`broadcast_tensors` but for shapes. This is equivalent to ``torch.broadcast_tensors(*map(torch.empty, shapes))[0].shape`` but avoids the need create to intermediate tensors. This is useful for broadcasting tensors of common batch shape but different rightmost shape, e.g. to broadcast mean vectors with covariance matrices. Example:: >>> torch.broadcast_shapes((2,), (3, 1), (1, 1, 1)) torch.Size([1, 3, 2]) Args: \*shapes (torch.Size): Shapes of tensors. Returns: shape (torch.Size): A shape compatible with all input shapes. Raises: RuntimeError: If shapes are incompatible. r)guard_size_oblivious is_nested_intz1Trying to create tensor with negative dimension (z): ()z=Shape mismatch: objects cannot be broadcast to a single shapezMInput shapes should be of type ints, a tuple of ints, or a list of ints, got cpu)deviceN)torchjit is_tracing isinstanceintSymInttuplelistlen%torch.fx.experimental.symbolic_shapesr/r0range RuntimeErrorSizeno_gradzerosexpandrshape) shapesmax_lenrFsresultr/r0iscalarr*s r+rrPs&6 99   ! !E%#u||!455Q;GEE4=11J;Gw E%#u||!455%%//r2E ?B7Ax!|*OPUPXzY]^c^f]gghi %UX..)338L!H 199%0A >>BV!H 1CC%yA~*[!&FI588#cAHzz&!!]]_[[E2F9?@}}U+G@'1G1:## _@_sH32H. H3.H33 Itensorsplit_size_or_sectionsdimreturn.cf[U5(a[[U4XUS9$URX5$)aSplits the tensor into chunks. Each chunk is a view of the original tensor. If :attr:`split_size_or_sections` is an integer type, then :attr:`tensor` will be split into equally sized chunks (if possible). Last chunk will be smaller if the tensor size along the given dimension :attr:`dim` is not divisible by :attr:`split_size`. If :attr:`split_size_or_sections` is a list, then :attr:`tensor` will be split into ``len(split_size_or_sections)`` chunks with sizes in :attr:`dim` according to :attr:`split_size_or_sections`. Args: tensor (Tensor): tensor to split. split_size_or_sections (int) or (list(int)): size of a single chunk or list of sizes for each chunk dim (int): dimension along which to split the tensor. Example:: >>> a = torch.arange(10).reshape(5, 2) >>> a tensor([[0, 1], [2, 3], [4, 5], [6, 7], [8, 9]]) >>> torch.split(a, 2) (tensor([[0, 1], [2, 3]]), tensor([[4, 5], [6, 7]]), tensor([[8, 9]])) >>> torch.split(a, [1, 4]) (tensor([[0, 1]]), tensor([[2, 3], [4, 5], [6, 7], [8, 9]])) rO)rrr")rMrNrOs r+r"r"s:X ''$ F9f#   <<. 44r,argsc^SSKJs Jn [U5S:a [ S5eSnSn[ US[ R5(azS[S[4SjmSRU4S jUS SS255n[U5S-S :Xa-US S RU4S jUS55-- nUSSS2nOUSSS2nO USnUS Sn[U5(a[[X2/UQ76$[U5S :Xa/[ US[[45(aUSn[U/UQ76$[U5S::dUR (d["R"X#5$SnUR%5(aTUR'5nUR("U/UQ7SUR*06Sn/[,R.R1U5Qn["R"X#US9$)aYeinsum(equation, *operands) -> Tensor Sums the product of the elements of the input :attr:`operands` along dimensions specified using a notation based on the Einstein summation convention. Einsum allows computing many common multi-dimensional linear algebraic array operations by representing them in a short-hand format based on the Einstein summation convention, given by :attr:`equation`. The details of this format are described below, but the general idea is to label every dimension of the input :attr:`operands` with some subscript and define which subscripts are part of the output. The output is then computed by summing the product of the elements of the :attr:`operands` along the dimensions whose subscripts are not part of the output. For example, matrix multiplication can be computed using einsum as `torch.einsum("ij,jk->ik", A, B)`. Here, j is the summation subscript and i and k the output subscripts (see section below for more details on why). Equation: The :attr:`equation` string specifies the subscripts (letters in `[a-zA-Z]`) for each dimension of the input :attr:`operands` in the same order as the dimensions, separating subscripts for each operand by a comma (','), e.g. `'ij,jk'` specify subscripts for two 2D operands. The dimensions labeled with the same subscript must be broadcastable, that is, their size must either match or be `1`. The exception is if a subscript is repeated for the same input operand, in which case the dimensions labeled with this subscript for this operand must match in size and the operand will be replaced by its diagonal along these dimensions. The subscripts that appear exactly once in the :attr:`equation` will be part of the output, sorted in increasing alphabetical order. The output is computed by multiplying the input :attr:`operands` element-wise, with their dimensions aligned based on the subscripts, and then summing out the dimensions whose subscripts are not part of the output. Optionally, the output subscripts can be explicitly defined by adding an arrow ('->') at the end of the equation followed by the subscripts for the output. For instance, the following equation computes the transpose of a matrix multiplication: 'ij,jk->ki'. The output subscripts must appear at least once for some input operand and at most once for the output. Ellipsis ('...') can be used in place of subscripts to broadcast the dimensions covered by the ellipsis. Each input operand may contain at most one ellipsis which will cover the dimensions not covered by subscripts, e.g. for an input operand with 5 dimensions, the ellipsis in the equation `'ab...c'` cover the third and fourth dimensions. The ellipsis does not need to cover the same number of dimensions across the :attr:`operands` but the 'shape' of the ellipsis (the size of the dimensions covered by them) must broadcast together. If the output is not explicitly defined with the arrow ('->') notation, the ellipsis will come first in the output (left-most dimensions), before the subscript labels that appear exactly once for the input operands. e.g. the following equation implements batch matrix multiplication `'...ij,...jk'`. A few final notes: the equation may contain whitespaces between the different elements (subscripts, ellipsis, arrow and comma) but something like `'. . .'` is not valid. An empty string `''` is valid for scalar operands. .. note:: ``torch.einsum`` handles ellipsis ('...') differently from NumPy in that it allows dimensions covered by the ellipsis to be summed over, that is, ellipsis are not required to be part of the output. .. note:: Please install opt-einsum (https://optimized-einsum.readthedocs.io/en/stable/) in order to enroll into a more performant einsum. You can install when installing torch like so: `pip install torch[opt-einsum]` or by itself with `pip install opt-einsum`. If opt-einsum is available, this function will automatically speed up computation and/or consume less memory by optimizing contraction order through our opt_einsum backend :mod:`torch.backends.opt_einsum` (The _ vs - is confusing, I know). This optimization occurs when there are at least three inputs, since the order does not matter otherwise. Note that finding `the` optimal path is an NP-hard problem, thus, opt-einsum relies on different heuristics to achieve near-optimal results. If opt-einsum is not available, the default order is to contract from left to right. To bypass this default behavior, add the following to disable opt_einsum and skip path calculation: ``torch.backends.opt_einsum.enabled = False`` To specify which strategy you'd like for opt_einsum to compute the contraction path, add the following line: ``torch.backends.opt_einsum.strategy = 'auto'``. The default strategy is 'auto', and we also support 'greedy' and 'optimal'. Disclaimer that the runtime of 'optimal' is factorial in the number of inputs! See more details in the opt_einsum documentation (https://optimized-einsum.readthedocs.io/en/stable/path_finding.html). .. note:: As of PyTorch 1.10 :func:`torch.einsum` also supports the sublist format (see examples below). In this format, subscripts for each operand are specified by sublists, list of integers in the range [0, 52). These sublists follow their operands, and an extra sublist can appear at the end of the input to specify the output's subscripts., e.g. `torch.einsum(op1, sublist1, op2, sublist2, ..., [subslist_out])`. Python's `Ellipsis` object may be provided in a sublist to enable broadcasting as described in the Equation section above. Args: equation (str): The subscripts for the Einstein summation. operands (List[Tensor]): The tensors to compute the Einstein summation of. Examples:: >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> # trace >>> torch.einsum('ii', torch.randn(4, 4)) tensor(-1.2104) >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> # diagonal >>> torch.einsum('ii->i', torch.randn(4, 4)) tensor([-0.1034, 0.7952, -0.2433, 0.4545]) >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> # outer product >>> x = torch.randn(5) >>> y = torch.randn(4) >>> torch.einsum('i,j->ij', x, y) tensor([[ 0.1156, -0.2897, -0.3918, 0.4963], [-0.3744, 0.9381, 1.2685, -1.6070], [ 0.7208, -1.8058, -2.4419, 3.0936], [ 0.1713, -0.4291, -0.5802, 0.7350], [ 0.5704, -1.4290, -1.9323, 2.4480]]) >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> # batch matrix multiplication >>> As = torch.randn(3, 2, 5) >>> Bs = torch.randn(3, 5, 4) >>> torch.einsum('bij,bjk->bik', As, Bs) tensor([[[-1.0564, -1.5904, 3.2023, 3.1271], [-1.6706, -0.8097, -0.8025, -2.1183]], [[ 4.2239, 0.3107, -0.5756, -0.2354], [-1.4558, -0.3460, 1.5087, -0.8530]], [[ 2.8153, 1.8787, -4.3839, -1.2112], [ 0.3728, -2.1131, 0.0921, 0.8305]]]) >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> # with sublist format and ellipsis >>> torch.einsum(As, [..., 0, 1], Bs, [..., 1, 2], [..., 0, 2]) tensor([[[-1.0564, -1.5904, 3.2023, 3.1271], [-1.6706, -0.8097, -0.8025, -2.1183]], [[ 4.2239, 0.3107, -0.5756, -0.2354], [-1.4558, -0.3460, 1.5087, -0.8530]], [[ 2.8153, 1.8787, -4.3839, -1.2112], [ 0.3728, -2.1131, 0.0921, 0.8305]]]) >>> # batch permute >>> A = torch.randn(2, 3, 4, 5) >>> torch.einsum('...ij->...ji', A).shape torch.Size([2, 3, 5, 4]) >>> # equivalent to torch.nn.functional.bilinear >>> A = torch.randn(3, 5, 4) >>> l = torch.randn(2, 5) >>> r = torch.randn(2, 4) >>> torch.einsum('bn,anm,bm->ba', l, A, r) tensor([[-0.3430, -5.2405, 0.4494], [ 0.3311, 5.5201, -3.0356]]) rNzteinsum(): must specify the equation string and at least one operand, or at least one operand and its subscripts listnrPcU[:XagUS:aUS:a[[S5U-5$US:a US:a[[S5U-S- 5$[S5e)Nz...rA4azKeinsum(): subscript in subscript list is not within the valid range [0, 52))Ellipsischrord ValueError)rVs r+parse_subscripteinsum..parse_subscriptsaH}Av!b&3s8a<((Bw1r63s8a<",--] r,,c3Z># UH nSRU4SjU55v M" g7f)c34># UH nT"U5v M g7fNr3.0rIr`s r+ #einsum...s#B1OA$6$6N)join)rhlr`s r+rieinsum..s#WJqBGG#B#BBBJs(+r.z->rdc34># UH nT"U5v M g7frfr3rgs r+rirns&L8aq'9'98rkr1optimize)path)torch.backends.opt_einsumbackends opt_einsumr>r_r9r6r r:strrlrrrr=r<enabledr is_availableget_opt_einsum contract_pathstrategy itertoolschain from_iterable) rSrtequationoperands _operandsrq _opt_einsum tupled_pathr`s @r+rrs^32 4y1} >  HH$q'5<<(( s s 88WDQRQUTUQUJWW t9q=A  rww&L48&LLL LHERE{HCaCyH78(##$VXK(KK 8}j!tUmDDQK h+++ 8}!3!3zz(-- D   //1 !//   *4*=*=  =..{;< ::ht 44r,)indexingr*rc[USU06$)Nr _meshgridrr*s r+r!r!s'5H55r,c[USU06$)aD Creates grids of coordinates specified by the 1D inputs in `attr`:tensors. This is helpful when you want to visualize data over some range of inputs. See below for a plotting example. Given :math:`N` 1D tensors :math:`T_0 \ldots T_{N-1}` as inputs with corresponding sizes :math:`S_0 \ldots S_{N-1}`, this creates :math:`N` N-dimensional tensors :math:`G_0 \ldots G_{N-1}`, each with shape :math:`(S_0, ..., S_{N-1})` where the output :math:`G_i` is constructed by expanding :math:`T_i` to the result shape. .. note:: 0D inputs are treated equivalently to 1D inputs of a single element. .. warning:: `torch.meshgrid(*tensors)` currently has the same behavior as calling `numpy.meshgrid(*arrays, indexing='ij')`. In the future `torch.meshgrid` will transition to `indexing='xy'` as the default. https://github.com/pytorch/pytorch/issues/50276 tracks this issue with the goal of migrating to NumPy's behavior. .. seealso:: :func:`torch.cartesian_prod` has the same effect but it collects the data in a tensor of vectors. Args: tensors (list of Tensor): list of scalars or 1 dimensional tensors. Scalars will be treated as tensors of size :math:`(1,)` automatically indexing: (str, optional): the indexing mode, either "xy" or "ij", defaults to "ij". See warning for future changes. If "xy" is selected, the first dimension corresponds to the cardinality of the second input and the second dimension corresponds to the cardinality of the first input. If "ij" is selected, the dimensions are in the same order as the cardinality of the inputs. Returns: seq (sequence of Tensors): If the input has :math:`N` tensors of size :math:`S_0 \ldots S_{N-1}``, then the output will also have :math:`N` tensors, where each tensor is of shape :math:`(S_0, ..., S_{N-1})`. Example:: >>> x = torch.tensor([1, 2, 3]) >>> y = torch.tensor([4, 5, 6]) Observe the element-wise pairings across the grid, (1, 4), (1, 5), ..., (3, 6). This is the same thing as the cartesian product. >>> grid_x, grid_y = torch.meshgrid(x, y, indexing='ij') >>> grid_x tensor([[1, 1, 1], [2, 2, 2], [3, 3, 3]]) >>> grid_y tensor([[4, 5, 6], [4, 5, 6], [4, 5, 6]]) This correspondence can be seen when these grids are stacked properly. >>> torch.equal(torch.cat(tuple(torch.dstack([grid_x, grid_y]))), ... torch.cartesian_prod(x, y)) True `torch.meshgrid` is commonly used to produce a grid for plotting. >>> # xdoctest: +REQUIRES(module:matplotlib) >>> # xdoctest: +REQUIRES(env:DOCTEST_SHOW) >>> import matplotlib.pyplot as plt >>> xs = torch.linspace(-5, 5, steps=100) >>> ys = torch.linspace(-5, 5, steps=100) >>> x, y = torch.meshgrid(xs, ys, indexing='xy') >>> z = torch.sin(torch.sqrt(x * x + y * y)) >>> ax = plt.axes(projection='3d') >>> ax.plot_surface(x.numpy(), y.numpy(), z.numpy()) >>> plt.show() .. image:: ../_static/img/meshgrid.png :width: 512 rrrs r+r!r!s|'5H55r,c[U5(a[[U/UQ7SU06$[U5S:Xa#[ US[ [ 45(aUSnUc0OSU0n[R"U40UD6$)Nrr.r)rrr!r>r9r=r<r)rr*kwargss r+rrst'""$XwTT8TT 7|qZ T5MBB!* #R*h)?F << *6 **r,Finputn_fft hop_length win_lengthwindowcenterpad_mode normalizedonesidedreturn_complexalign_to_windowc [U5(a[[U4UUUUUUUUUU U S9 $U(aU b [S5eU(aUR 5n S/SU - -[ UR 55-n [US-5n [R"URU 5X/U5nURURU *S5n[R"UUUUUUUU U 5 $)aShort-time Fourier transform (STFT). .. warning:: From version 1.8.0, :attr:`return_complex` must always be given explicitly for real inputs and `return_complex=False` has been deprecated. Strongly prefer `return_complex=True` as in a future pytorch release, this function will only return complex tensors. Note that :func:`torch.view_as_real` can be used to recover a real tensor with an extra last dimension for real and imaginary components. .. warning:: From version 2.1, a warning will be provided if a :attr:`window` is not specified. In a future release, this attribute will be required. Not providing a window currently defaults to using a rectangular window, which may result in undesirable artifacts. Consider using tapered windows, such as :func:`torch.hann_window`. The STFT computes the Fourier transform of short overlapping windows of the input. This giving frequency components of the signal as they change over time. The interface of this function is modeled after (but *not* a drop-in replacement for) librosa_ stft function. .. _librosa: https://librosa.org/doc/latest/generated/librosa.stft.html Ignoring the optional batch dimension, this method computes the following expression: .. math:: X[\omega, m] = \sum_{k = 0}^{\text{win\_length-1}}% \text{window}[k]\ \text{input}[m \times \text{hop\_length} + k]\ % \exp\left(- j \frac{2 \pi \cdot \omega k}{\text{n\_fft}}\right), where :math:`m` is the index of the sliding window, and :math:`\omega` is the frequency :math:`0 \leq \omega < \text{n\_fft}` for ``onesided=False``, or :math:`0 \leq \omega < \lfloor \text{n\_fft} / 2 \rfloor + 1` for ``onesided=True``. * :attr:`input` must be either a 1-D time sequence or a 2-D batch of time sequences. * If :attr:`hop_length` is ``None`` (default), it is treated as equal to ``floor(n_fft / 4)``. * If :attr:`win_length` is ``None`` (default), it is treated as equal to :attr:`n_fft`. * :attr:`window` can be a 1-D tensor of size :attr:`win_length`, e.g., from :meth:`torch.hann_window`. If :attr:`window` is ``None`` (default), it is treated as if having :math:`1` everywhere in the window. If :math:`\text{win\_length} < \text{n\_fft}`, :attr:`window` will be padded on both sides to length :attr:`n_fft` before being applied. * If :attr:`center` is ``True`` (default), :attr:`input` will be padded on both sides so that the :math:`t`-th frame is centered at time :math:`t \times \text{hop\_length}`. Otherwise, the :math:`t`-th frame begins at time :math:`t \times \text{hop\_length}`. * :attr:`pad_mode` determines the padding method used on :attr:`input` when :attr:`center` is ``True``. See :meth:`torch.nn.functional.pad` for all available options. Default is ``"reflect"``. * If :attr:`onesided` is ``True`` (default for real input), only values for :math:`\omega` in :math:`\left[0, 1, 2, \dots, \left\lfloor \frac{\text{n\_fft}}{2} \right\rfloor + 1\right]` are returned because the real-to-complex Fourier transform satisfies the conjugate symmetry, i.e., :math:`X[m, \omega] = X[m, \text{n\_fft} - \omega]^*`. Note if the input or window tensors are complex, then :attr:`onesided` output is not possible. * If :attr:`normalized` is ``True`` (default is ``False``), the function returns the normalized STFT results, i.e., multiplied by :math:`(\text{frame\_length})^{-0.5}`. * If :attr:`return_complex` is ``True`` (default if input is complex), the return is a ``input.dim() + 1`` dimensional complex tensor. If ``False``, the output is a ``input.dim() + 2`` dimensional real tensor where the last dimension represents the real and imaginary components. Returns either a complex tensor of size :math:`(* \times N \times T)` if :attr:`return_complex` is true, or a real tensor of size :math:`(* \times N \times T \times 2)`. Where :math:`*` is the optional batch size of :attr:`input`, :math:`N` is the number of frequencies where STFT is applied and :math:`T` is the total number of frames used. .. warning:: This function changed signature at version 0.4.1. Calling with the previous signature may cause error or return incorrect result. Args: input (Tensor): the input tensor of shape `(B?, L)` where `B?` is an optional batch dimension n_fft (int): size of Fourier transform hop_length (int, optional): the distance between neighboring sliding window frames. Default: ``None`` (treated as equal to ``floor(n_fft / 4)``) win_length (int, optional): the size of window frame and STFT filter. Default: ``None`` (treated as equal to :attr:`n_fft`) window (Tensor, optional): the optional window function. Shape must be 1d and `<= n_fft` Default: ``None`` (treated as window of all :math:`1` s) center (bool, optional): whether to pad :attr:`input` on both sides so that the :math:`t`-th frame is centered at time :math:`t \times \text{hop\_length}`. Default: ``True`` pad_mode (str, optional): controls the padding method used when :attr:`center` is ``True``. Default: ``"reflect"`` normalized (bool, optional): controls whether to return the normalized STFT results Default: ``False`` onesided (bool, optional): controls whether to return half of results to avoid redundancy for real inputs. Default: ``True`` for real :attr:`input` and :attr:`window`, ``False`` otherwise. return_complex (bool, optional): whether to return a complex tensor, or a real tensor with an extra last dimension for the real and imaginary components. .. versionchanged:: 2.0 ``return_complex`` is now a required argument for real inputs, as the default is being transitioned to ``True``. .. deprecated:: 2.0 ``return_complex=False`` is deprecated, instead use ``return_complex=True`` Note that calling :func:`torch.view_as_real` on the output will recover the deprecated output format. Returns: Tensor: A tensor containing the STFT result with shape `(B?, N, T, C?)` where - `B?` is an optional batch dimension from the input. - `N` is the number of frequency samples, `(n_fft // 2) + 1` for `onesided=True`, or otherwise `n_fft`. - `T` is the number of frames, `1 + L // hop_length` for `center=True`, or `1 + (L - n_fft) // hop_length` otherwise. - `C?` is an optional length-2 dimension of real and imaginary components, present when `return_complex=False`. ) rrrrrrrrrNz;stft align_to_window should only be set when center = falser.rU) rrr#rArOr=sizer:FpadviewrFr)rrrrrrrrrrr signal_dimextended_shapers r+r#r#-sb &&$  H  !!!)+  /- I  YY[ J/$uzz|2DD%1*oejj03*hG 5;; {|45 88    r,aistft(input, n_fft, hop_length=None, win_length=None, window=None, center=True, normalized=False, onesided=None, length=None, return_complex=False) -> Tensor: Inverse short time Fourier Transform. This is expected to be the inverse of :func:`~torch.stft`. .. warning:: From version 2.1, a warning will be provided if a :attr:`window` is not specified. In a future release, this attribute will be required. Please provide the same window used in the stft call. It has the same parameters (+ additional optional parameter of :attr:`length`) and it should return the least squares estimation of the original signal. The algorithm will check using the NOLA condition ( nonzero overlap). Important consideration in the parameters :attr:`window` and :attr:`center` so that the envelope created by the summation of all the windows is never zero at certain point in time. Specifically, :math:`\sum_{t=-\infty}^{\infty} |w|^2[n-t\times hop\_length] \cancel{=} 0`. Since :func:`~torch.stft` discards elements at the end of the signal if they do not fit in a frame, ``istft`` may return a shorter signal than the original signal (can occur if :attr:`center` is False since the signal isn't padded). If `length` is given in the arguments and is longer than expected, ``istft`` will pad zeros to the end of the returned signal. If :attr:`center` is ``True``, then there will be padding e.g. ``'constant'``, ``'reflect'``, etc. Left padding can be trimmed off exactly because they can be calculated but right padding cannot be calculated without additional information. Example: Suppose the last window is: ``[17, 18, 0, 0, 0]`` vs ``[18, 0, 0, 0, 0]`` The :attr:`n_fft`, :attr:`hop_length`, :attr:`win_length` are all the same which prevents the calculation of right padding. These additional values could be zeros or a reflection of the signal so providing :attr:`length` could be useful. If :attr:`length` is ``None`` then padding will be aggressively removed (some loss of signal). [1] D. W. Griffin and J. S. Lim, "Signal estimation from modified short-time Fourier transform," IEEE Trans. ASSP, vol.32, no.2, pp.236-243, Apr. 1984. Args: input (Tensor): The input tensor. Expected to be in the format of :func:`~torch.stft`, output. That is a complex tensor of shape `(B?, N, T)` where - `B?` is an optional batch dimension - `N` is the number of frequency samples, `(n_fft // 2) + 1` for onesided input, or otherwise `n_fft`. - `T` is the number of frames, `1 + length // hop_length` for centered stft, or `1 + (length - n_fft) // hop_length` otherwise. .. versionchanged:: 2.0 Real datatype inputs are no longer supported. Input must now have a complex datatype, as returned by ``stft(..., return_complex=True)``. n_fft (int): Size of Fourier transform hop_length (Optional[int]): The distance between neighboring sliding window frames. (Default: ``n_fft // 4``) win_length (Optional[int]): The size of window frame and STFT filter. (Default: ``n_fft``) window (Optional[torch.Tensor]): The optional window function. Shape must be 1d and `<= n_fft` (Default: ``torch.ones(win_length)``) center (bool): Whether :attr:`input` was padded on both sides so that the :math:`t`-th frame is centered at time :math:`t \times \text{hop\_length}`. (Default: ``True``) normalized (bool): Whether the STFT was normalized. (Default: ``False``) onesided (Optional[bool]): Whether the STFT was onesided. (Default: ``True`` if `n_fft != fft_size` in the input size) length (Optional[int]): The amount to trim the signal by (i.e. the original signal length). Defaults to `(T - 1) * hop_length` for centered stft, or `n_fft + (T - 1) * hop_length` otherwise, where `T` is the number of input frames. return_complex (Optional[bool]): Whether the output should be complex, or if the input should be assumed to derive from a real signal and window. Note that this is incompatible with ``onesided=True``. (Default: ``False``) Returns: Tensor: Least squares estimation of the original signal of shape `(B?, length)` where `B?` is an optional batch dimension from the input tensor. sortedreturn_inverse return_countsc [U5(a[[U4UUUUUS9$Ub[R"UUUUUS9upVnO[ R "UUUUS9upVnXVU4$)aunique(input, sorted=True, return_inverse=False, return_counts=False, dim=None) -> tuple[Tensor, Tensor, Tensor] Returns the unique elements of the input tensor. .. note:: This function is different from :func:`torch.unique_consecutive` in the sense that this function also eliminates non-consecutive duplicate values. .. note:: Currently in the CUDA implementation and the CPU implementation, `torch.unique` always sort the tensor at the beginning regardless of the `sort` argument. Sorting could be slow, so if your input tensor is already sorted, it is recommended to use :func:`torch.unique_consecutive` which avoids the sorting. Args: input (Tensor): the input tensor sorted (bool): Whether to sort the unique elements in ascending order before returning as output. return_inverse (bool): Whether to also return the indices for where elements in the original input ended up in the returned unique list. return_counts (bool): Whether to also return the counts for each unique element. dim (int, optional): the dimension to operate upon. If ``None``, the unique of the flattened input is returned. Otherwise, each of the tensors indexed by the given dimension is treated as one of the elements to apply the unique operation upon. See examples for more details. Default: ``None`` Returns: (Tensor, Tensor (optional), Tensor (optional)): A tensor or a tuple of tensors containing - **output** (*Tensor*): the output list of unique scalar elements. - **inverse_indices** (*Tensor*): (optional) if :attr:`return_inverse` is True, there will be an additional returned tensor (same shape as input) representing the indices for where elements in the original input map to in the output; otherwise, this function will only return a single tensor. - **counts** (*Tensor*): (optional) if :attr:`return_counts` is True, there will be an additional returned tensor (same shape as output or output.size(dim), if dim was specified) representing the number of occurrences for each unique value or tensor. Example:: >>> output = torch.unique(torch.tensor([1, 3, 2, 3], dtype=torch.long)) >>> output tensor([1, 2, 3]) >>> output, inverse_indices = torch.unique( ... torch.tensor([1, 3, 2, 3], dtype=torch.long), sorted=True, return_inverse=True) >>> output tensor([1, 2, 3]) >>> inverse_indices tensor([0, 2, 1, 2]) >>> output, inverse_indices = torch.unique( ... torch.tensor([[1, 3], [2, 3]], dtype=torch.long), sorted=True, return_inverse=True) >>> output tensor([1, 2, 3]) >>> inverse_indices tensor([[0, 2], [1, 2]]) >>> a = torch.tensor([ ... [ ... [1, 1, 0, 0], ... [1, 1, 0, 0], ... [0, 0, 1, 1], ... ], ... [ ... [0, 0, 1, 1], ... [0, 0, 1, 1], ... [1, 1, 1, 1], ... ], ... [ ... [1, 1, 0, 0], ... [1, 1, 0, 0], ... [0, 0, 1, 1], ... ], ... ]) >>> # If we call `torch.unique(a, dim=0)`, each of the tensors `a[idx, :, :]` >>> # will be compared. We can see that `a[0, :, :]` and `a[2, :, :]` match >>> # each other, so one of them will be removed. >>> (a[0, :, :] == a[2, :, :]).all() tensor(True) >>> a_unique_dim0 = torch.unique(a, dim=0) >>> a_unique_dim0 tensor([[[0, 0, 1, 1], [0, 0, 1, 1], [1, 1, 1, 1]], [[1, 1, 0, 0], [1, 1, 0, 0], [0, 0, 1, 1]]]) >>> # Notice which sub-tensors from `a` match with the sub-tensors from >>> # `a_unique_dim0`: >>> (a_unique_dim0[0, :, :] == a[1, :, :]).all() tensor(True) >>> (a_unique_dim0[1, :, :] == a[0, :, :]).all() tensor(True) >>> # For `torch.unique(a, dim=1)`, each of the tensors `a[:, idx, :]` are >>> # compared. `a[:, 0, :]` and `a[:, 1, :]` match each other, so one of >>> # them will be removed. >>> (a[:, 0, :] == a[:, 1, :]).all() tensor(True) >>> torch.unique(a, dim=1) tensor([[[0, 0, 1, 1], [1, 1, 0, 0]], [[1, 1, 1, 1], [0, 0, 1, 1]], [[0, 0, 1, 1], [1, 1, 0, 0]]]) >>> # For `torch.unique(a, dim=2)`, the tensors `a[:, :, idx]` are compared. >>> # `a[:, :, 0]` and `a[:, :, 1]` match each other. Also, `a[:, :, 2]` and >>> # `a[:, :, 3]` match each other as well. So in this case, two of the >>> # sub-tensors will be removed. >>> (a[:, :, 0] == a[:, :, 1]).all() tensor(True) >>> (a[:, :, 2] == a[:, :, 3]).all() tensor(True) >>> torch.unique(a, dim=2) tensor([[[0, 1], [0, 1], [1, 0]], [[1, 0], [1, 0], [1, 1]], [[0, 1], [0, 1], [1, 0]]]) )rrrrO)rrr)rrr%r unique_dimr6_unique2)rrrrrOoutputinverse_indicescountss r+ _unique_implrCsX &&$  H )'   *-..  )' + '+0.. )' + ' F **r,c [U5(a[[U4UUUUS9$[R"XX#S9upEnXEU4$)aEliminates all but the first element from every consecutive group of equivalent elements. .. note:: This function is different from :func:`torch.unique` in the sense that this function only eliminates consecutive duplicate values. This semantics is similar to `std::unique` in C++. Args: input (Tensor): the input tensor return_inverse (bool): Whether to also return the indices for where elements in the original input ended up in the returned unique list. return_counts (bool): Whether to also return the counts for each unique element. dim (int): the dimension to apply unique. If ``None``, the unique of the flattened input is returned. default: ``None`` Returns: (Tensor, Tensor (optional), Tensor (optional)): A tensor or a tuple of tensors containing - **output** (*Tensor*): the output list of unique scalar elements. - **inverse_indices** (*Tensor*): (optional) if :attr:`return_inverse` is True, there will be an additional returned tensor (same shape as input) representing the indices for where elements in the original input map to in the output; otherwise, this function will only return a single tensor. - **counts** (*Tensor*): (optional) if :attr:`return_counts` is True, there will be an additional returned tensor (same shape as output or output.size(dim), if dim was specified) representing the number of occurrences for each unique value or tensor. Example:: >>> x = torch.tensor([1, 1, 2, 2, 3, 1, 1, 2]) >>> output = torch.unique_consecutive(x) >>> output tensor([1, 2, 3, 1, 2]) >>> output, inverse_indices = torch.unique_consecutive(x, return_inverse=True) >>> output tensor([1, 2, 3, 1, 2]) >>> inverse_indices tensor([0, 0, 1, 1, 2, 3, 3, 4]) >>> output, counts = torch.unique_consecutive(x, return_counts=True) >>> output tensor([1, 2, 3, 1, 2]) >>> counts tensor([2, 2, 1, 2, 1]) )rrrO)rrr&r)rrrrOrrrs r+_unique_consecutive_implrsZn &&$  H )'   '*&<&< M'#FV F **r,cb[U5(a [XX#U5$[XX#U5upVnXW4$rfrr)rrrrrOr_rs r+_return_countsr2s: &&E>#NN$UNSVWFv >r,cb[U5(a [XX#U5$[XX#U5un nU$rfr)rrrrrOrrs r+_return_outputrBs7 &&E>#NN~cRLFAq Mr,cb[U5(a [XX#U5$[XX#U5upVnXV4$rfr)rrrrrOrrrs r+_return_inverserRs? &&E>#NN!- ~c"FQ  ""r,rr%)arg_name arg_indexdefaultif_trueif_false module_name func_namerUc^[U5(a [XX#5$[XX#5upEnXF4$rfrr)rrrrOrrrs r+_consecutive_return_countsrs: &&'}RR0 }Fv >r,c^[U5(a [XX#5$[XX#5un nU$rfr)rrrrOrrs r+_consecutive_return_outputrs3 &&'}RR+E=VLFAq Mr,c^[U5(a [XX#5$[XX#5upEnXE4$rfr)rrrrOrrrs r+_consecutive_return_inversers; &&'}RR!9 }"FQ  ""r,r.r&dimsoutcgrfr3r[brrs r+r$r$ r,cgrfr3rs r+r$r$rr,cgrfr3rs r+r$r$rr,cgrfr3rs r+r$r$rr,c [X5(a[[X4XX#S9$[U[[ [ R[[ R45(d[SSU3-5e/n/n[U[[ 45(aUupE[U[ R5(aUR5nUS:aUR5SS:Xde[ RR[ [USR55n[ RR[ [USR55nOW[UR!55nUS:a[SU35e[ [#U*S55n[ [#U55n[U[[ R45(axUS:a[SU35eU[%UR'5UR'55:a[SU35e[ [#U*S55n[ [#U55nUc[(R"XXE5$[(R"XXEUS 9$) aReturns a contraction of a and b over multiple dimensions. :attr:`tensordot` implements a generalized matrix product. Args: a (Tensor): Left tensor to contract b (Tensor): Right tensor to contract dims (int or Tuple[List[int], List[int]] or List[List[int]] containing two lists or Tensor): number of dimensions to contract or explicit lists of dimensions for :attr:`a` and :attr:`b` respectively When called with a non-negative integer argument :attr:`dims` = :math:`d`, and the number of dimensions of :attr:`a` and :attr:`b` is :math:`m` and :math:`n`, respectively, :func:`~torch.tensordot` computes .. math:: r_{i_0,...,i_{m-d}, i_d,...,i_n} = \sum_{k_0,...,k_{d-1}} a_{i_0,...,i_{m-d},k_0,...,k_{d-1}} \times b_{k_0,...,k_{d-1}, i_d,...,i_n}. When called with :attr:`dims` of the list form, the given dimensions will be contracted in place of the last :math:`d` of :attr:`a` and the first :math:`d` of :math:`b`. The sizes in these dimensions must match, but :func:`~torch.tensordot` will deal with broadcasted dimensions. Examples:: >>> a = torch.arange(60.).reshape(3, 4, 5) >>> b = torch.arange(24.).reshape(4, 3, 2) >>> torch.tensordot(a, b, dims=([1, 0], [0, 1])) tensor([[4400., 4730.], [4532., 4874.], [4664., 5018.], [4796., 5162.], [4928., 5306.]]) >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA) >>> a = torch.randn(3, 4, 5, device='cuda') >>> b = torch.randn(4, 5, 6, device='cuda') >>> c = torch.tensordot(a, b, dims=2).cpu() tensor([[ 8.3504, -2.5436, 6.2922, 2.7556, -1.0732, 3.2741], [ 3.3161, 0.0704, 5.0187, -0.4079, -4.3126, 4.8744], [ 0.8223, 3.9445, 3.2168, -0.2400, 3.4117, 1.7780]]) >>> a = torch.randn(3, 5, 4, 6) >>> b = torch.randn(6, 4, 5, 3) >>> torch.tensordot(a, b, dims=([2, 1, 3], [1, 2, 0])) tensor([[ 7.7193, -2.4867, -10.3204], [ 1.5513, -14.4737, -6.5113], [ -0.2850, 4.2573, -3.5997]]) )rrzqtensordot expects dims to be int or tuple[list[int], list[int]] or list[list[int]] containing two lists, but got zdims=r.rrUz*tensordot expects dims >= 0, but got dims=z8tensordot expects dims < ndim_a or ndim_b, but got dims=r)rrr$r9r<r=r6r r:r;rAnumelrr7annotatetolistitemr@minrOr)r[rrrdims_adims_b num_elementsdims_vals r+r$r$sp#1(($Y4QQ dUD%,,U\\J K K ?dVn   FF$ &&$ %%zz| ! 99;q>Q& &&YY''S 47>>3CDFYY''S 47>>3CDF499;'H!|"%OPTv#VWW% 1-.F%/*F$ell+,, !8!KD6RS S #aeegquuw' 'J4&Q eTE1o&eDk" {}}Q622}}Q6s;;r,cp[U5(a[[U/UQ76$[R"U5$)aDo cartesian product of the given sequence of tensors. The behavior is similar to python's `itertools.product`. Args: *tensors: any number of 1 dimensional tensors. Returns: Tensor: A tensor equivalent to converting all the input tensors into lists, do `itertools.product` on these lists, and finally convert the resulting list into tensor. Example:: >>> import itertools >>> a = [1, 2, 3] >>> b = [4, 5] >>> list(itertools.product(a, b)) [(1, 4), (1, 5), (2, 4), (2, 5), (3, 4), (3, 5)] >>> tensor_a = torch.tensor(a) >>> tensor_b = torch.tensor(b) >>> torch.cartesian_prod(tensor_a, tensor_b) tensor([[1, 4], [1, 5], [2, 4], [2, 5], [3, 4], [3, 5]]) )rrrrr)s r+rrks2<'""$^WGwGG   g &&r,c[U5(a[[U/UQ76$[RR RU5$)aCreate a block diagonal matrix from provided tensors. Args: *tensors: One or more tensors with 0, 1, or 2 dimensions. Returns: Tensor: A 2 dimensional tensor with all the input tensors arranged in order such that their upper left and lower right corners are diagonally adjacent. All other elements are set to 0. Example:: >>> import torch >>> A = torch.tensor([[0, 1], [1, 0]]) >>> B = torch.tensor([[3, 4, 5], [6, 7, 8]]) >>> C = torch.tensor(7) >>> D = torch.tensor([1, 2, 3]) >>> E = torch.tensor([[4], [5], [6]]) >>> torch.block_diag(A, B, C, D, E) tensor([[0, 1, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 3, 4, 5, 0, 0, 0, 0, 0], [0, 0, 6, 7, 8, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 7, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 1, 2, 3, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 4], [0, 0, 0, 0, 0, 0, 0, 0, 0, 5], [0, 0, 0, 0, 0, 0, 0, 0, 0, 6]]) )rrrr6_C_VariableFunctionsr)s r+rrs<>'""$ZC7CC 88 & & 1 1' ::r,c [X5(a[[X4XX#S9$US:Xa[R"XUS5$US:Xa[R"XUS5$US:Xa[R"XUS5$[ US35e) aComputes batched the p-norm distance between each pair of the two collections of row vectors. Args: x1 (Tensor): input tensor of shape :math:`B \times P \times M`. x2 (Tensor): input tensor of shape :math:`B \times R \times M`. p: p value for the p-norm distance to calculate between each vector pair :math:`\in [0, \infty]`. compute_mode: 'use_mm_for_euclid_dist_if_necessary' - will use matrix multiplication approach to calculate euclidean distance (p = 2) if P > 25 or R > 25 'use_mm_for_euclid_dist' - will always use matrix multiplication approach to calculate euclidean distance (p = 2) 'donot_use_mm_for_euclid_dist' - will never use matrix multiplication approach to calculate euclidean distance (p = 2) Default: use_mm_for_euclid_dist_if_necessary. If x1 has shape :math:`B \times P \times M` and x2 has shape :math:`B \times R \times M` then the output will have shape :math:`B \times P \times R`. This function is equivalent to `scipy.spatial.distance.cdist(input,'minkowski', p=p)` if :math:`p \in (0, \infty)`. When :math:`p = 0` it is equivalent to `scipy.spatial.distance.cdist(input, 'hamming') * M`. When :math:`p = \infty`, the closest scipy function is `scipy.spatial.distance.cdist(xn, lambda x, y: np.abs(x - y).max())`. Example: >>> a = torch.tensor([[0.9041, 0.0196], [-0.3108, -2.4423], [-0.4821, 1.059]]) >>> a tensor([[ 0.9041, 0.0196], [-0.3108, -2.4423], [-0.4821, 1.0590]]) >>> b = torch.tensor([[-2.1763, -0.4713], [-0.6986, 1.3702]]) >>> b tensor([[-2.1763, -0.4713], [-0.6986, 1.3702]]) >>> torch.cdist(a, b, p=2) tensor([[3.1193, 2.0959], [2.7138, 3.8322], [2.2830, 0.3791]]) )p compute_mode#use_mm_for_euclid_dist_if_necessaryNuse_mm_for_euclid_distr.donot_use_mm_for_euclid_distrUz& is not a valid value for compute_mode)rrrrr_)x1x2rrs r+rrsT#2**$ B8Rq  <<yyD)) 1 1yyA&& 7 7yyA&&L>)OPQQr,c[U5(a[[U/UQ76$[U5S:XaUSn[R"U5$)a> Returns a 1-dimensional view of each input tensor with zero dimensions. Input tensors with one or more dimensions are returned as-is. Args: input (Tensor or list of Tensors) Returns: output (Tensor or tuple of Tensors) Example:: >>> x = torch.arange(2) >>> x tensor([0, 1]) >>> torch.atleast_1d(x) tensor([0, 1]) >>> x = torch.tensor(1.) >>> x tensor(1.) >>> torch.atleast_1d(x) tensor([1.]) >>> x = torch.tensor(0.5) >>> y = torch.tensor(1.) >>> torch.atleast_1d((x, y)) (tensor([0.5000]), tensor([1.])) r.r)rrrr>rr)s r+rrsE:'""$ZC7CC 7|q!* >>' ""r,c[U5(a[[U/UQ76$[U5S:XaUSn[R"U5$)a{ Returns a 2-dimensional view of each input tensor with zero dimensions. Input tensors with two or more dimensions are returned as-is. Args: input (Tensor or list of Tensors) Returns: output (Tensor or tuple of Tensors) Example:: >>> x = torch.tensor(1.) >>> x tensor(1.) >>> torch.atleast_2d(x) tensor([[1.]]) >>> x = torch.arange(4).view(2, 2) >>> x tensor([[0, 1], [2, 3]]) >>> torch.atleast_2d(x) tensor([[0, 1], [2, 3]]) >>> x = torch.tensor(0.5) >>> y = torch.tensor(1.) >>> torch.atleast_2d((x, y)) (tensor([[0.5000]]), tensor([[1.]])) r.r)rrrr>rr)s r+rrsE>'""$ZC7CC 7|q!* >>' ""r,c[U5(a[[U/UQ76$[U5S:XaUSn[R"U5$)a> Returns a 3-dimensional view of each input tensor with zero dimensions. Input tensors with three or more dimensions are returned as-is. Args: input (Tensor or list of Tensors) Returns: output (Tensor or tuple of Tensors) Example: >>> x = torch.tensor(0.5) >>> x tensor(0.5000) >>> torch.atleast_3d(x) tensor([[[0.5000]]]) >>> y = torch.arange(4).view(2, 2) >>> y tensor([[0, 1], [2, 3]]) >>> torch.atleast_3d(y) tensor([[[0], [1]], [[2], [3]]]) >>> x = torch.tensor(1).view(1, 1, 1) >>> x tensor([[[1]]]) >>> torch.atleast_3d(x) tensor([[[1]]]) >>> x = torch.tensor(0.5) >>> y = torch.tensor(1.0) >>> torch.atleast_3d((x, y)) (tensor([[[0.5000]]]), tensor([[[1.]]])) r.r)rrrr>rr)s r+rr4sFN'""$ZC7CC 7|q!* >>' ""r,cgrfr3rrrOkeepdimrdtypes r+r r q r,cgrfr3rs r+r r }rr,cgrfr3rs r+r r rr,cgrfr3rs r+r r rr,rc "[U5(a[[U4XX#XES9$UR[R :XGaUR RSSS[RRR4;GapUb,[U[[R45(aU/nOUnOSn[U[5(aUS:XazUb4[U[[R45(d[U5S::aCUc [R R#USXcUS9$[R R#USXcXTS 9$Uc[%['UR(55nUc[R R+XXcUS9$[R R+XXcXTS 9$UcS OUnUc[R R#XXcUS9$[R R#XXcXTS 9$UR-5nUcxUcuUcrUbo[U[5(aUS:Xa[.R0"US US 9$[U[5(d)[%['U55n[.R"XXcS 9$Ub,[U[[R45(aU/nOUnOSn[U[5(aUS:XaRUb [3S 5eUc[%['U55nUc[.R0"XUS9$[.R0"XX4S9$US:XajUb [3S5eUc,Uc[.R4"XS9$[.R4"XUS9$Uc[.R4"XUS9$[.R4"XX4S9$[7SU35eUc[%['U55nUc.Uc[.R"XXcS9$[.R"XXcUS9$Uc[.R"XXcUS9$[.R"XXcXTS9$)aReturns the matrix norm or vector norm of a given tensor. .. warning:: torch.norm is deprecated and may be removed in a future PyTorch release. Its documentation and behavior may be incorrect, and it is no longer actively maintained. Use :func:`torch.linalg.vector_norm` when computing vector norms and :func:`torch.linalg.matrix_norm` when computing matrix norms. For a function with a similar behavior as this one see :func:`torch.linalg.norm`. Note, however, the signature for these functions is slightly different than the signature for ``torch.norm``. Args: input (Tensor): The input tensor. Its data type must be either a floating point or complex type. For complex inputs, the norm is calculated using the absolute value of each element. If the input is complex and neither :attr:`dtype` nor :attr:`out` is specified, the result's data type will be the corresponding floating point type (e.g. float if :attr:`input` is complexfloat). p (int, float, inf, -inf, 'fro', 'nuc', optional): the order of norm. Default: ``'fro'`` The following norms can be calculated: ====== ============== ========================== ord matrix norm vector norm ====== ============== ========================== 'fro' Frobenius norm -- 'nuc' nuclear norm -- Number -- sum(abs(x)**ord)**(1./ord) ====== ============== ========================== The vector norm can be calculated across any number of dimensions. The corresponding dimensions of :attr:`input` are flattened into one dimension, and the norm is calculated on the flattened dimension. Frobenius norm produces the same result as ``p=2`` in all cases except when :attr:`dim` is a list of three or more dims, in which case Frobenius norm throws an error. Nuclear norm can only be calculated across exactly two dimensions. dim (int, tuple of ints, list of ints, optional): Specifies which dimension or dimensions of :attr:`input` to calculate the norm across. If :attr:`dim` is ``None``, the norm will be calculated across all dimensions of :attr:`input`. If the norm type indicated by :attr:`p` does not support the specified number of dimensions, an error will occur. keepdim (bool, optional): whether the output tensors have :attr:`dim` retained or not. Ignored if :attr:`dim` = ``None`` and :attr:`out` = ``None``. Default: ``False`` out (Tensor, optional): the output tensor. Ignored if :attr:`dim` = ``None`` and :attr:`out` = ``None``. dtype (:class:`torch.dtype`, optional): the desired data type of returned tensor. If specified, the input tensor is casted to :attr:`dtype` while performing the operation. Default: None. .. note:: Even though ``p='fro'`` supports any number of dimensions, the true mathematical definition of Frobenius norm only applies to tensors with exactly two dimensions. :func:`torch.linalg.matrix_norm` with ``ord='fro'`` aligns with the mathematical definition, since it can only be applied across exactly two dimensions. Example:: >>> import torch >>> a = torch.arange(9, dtype= torch.float) - 4 >>> b = a.reshape((3, 3)) >>> torch.norm(a) tensor(7.7460) >>> torch.norm(b) tensor(7.7460) >>> torch.norm(a, float('inf')) tensor(4.) >>> torch.norm(b, float('inf')) tensor(4.) >>> c = torch.tensor([[ 1, 2, 3], [-1, 1, 4]] , dtype=torch.float) >>> torch.norm(c, dim=0) tensor([1.4142, 2.2361, 5.0000]) >>> torch.norm(c, dim=1) tensor([3.7417, 4.2426]) >>> torch.norm(c, p=1, dim=1) tensor([6., 6.]) >>> d = torch.arange(8, dtype=torch.float).reshape(2, 2, 2) >>> torch.norm(d, dim=(1, 2)) tensor([ 3.7417, 11.2250]) >>> torch.norm(d[0, :, :]), torch.norm(d[1, :, :]) (tensor(3.7417), tensor(11.2250)) )rrOrrrr4cudametaNfrorUr)rr@r3)rOrz1dtype argument is not supported in frobenius norm)r)rrnucz/dtype argument is not supported in nuclear normz4only valid string values are 'fro' and 'nuc', found )rr)rrr)rrr layoutr6stridedr5typeutilsbackend_registration_privateuse1_backend_namer9r:r;rur>linalg vector_normr=r@ndim matrix_normrOrfrobenius_normr_ nuclear_normrA) rrrOrrr_dim_prs r+r r sJ &&$ 5(ECc   ||u}}$):):  ((BB ?* ?#U\\233uD a  Ez z#U\\/BCCs3xST}; <<33q$u4!<<33q$u4 |E%**-.{||//$u/UU||//d50  qB{||//4PU/VV||//tE0 99;D {s{u} a  Ez))%RII!S!!d $D88E$@ @   cC. / /5DD!S :  !TUU|E$K({))%wGG))%wPP %Z  !RSS|;++ECC++ELL;++EII++ERRQRSQTUVV <d $D ;}xx$@@xx$uMM}xx$SIIxx$uVVr,indicesrFcz[U5(a[[U4XS9$[X5nUR S5$)aConverts a tensor of flat indices into a tuple of coordinate tensors that index into an arbitrary tensor of the specified shape. Args: indices (Tensor): An integer tensor containing indices into the flattened version of an arbitrary tensor of shape :attr:`shape`. All elements must be in the range ``[0, prod(shape) - 1]``. shape (int, sequence of ints, or torch.Size): The shape of the arbitrary tensor. All elements must be non-negative. Returns: tuple of Tensors: Each ``i``-th tensor in the output corresponds with dimension ``i`` of :attr:`shape`. Each tensor has the same shape as ``indices`` and contains one index into dimension ``i`` for each of the flat indices given by ``indices``. Example:: >>> import torch >>> torch.unravel_index(torch.tensor(4), (3, 2)) (tensor(2), tensor(0)) >>> torch.unravel_index(torch.tensor([4, 1]), (3, 2)) (tensor([2, 0]), tensor([0, 1])) >>> torch.unravel_index(torch.tensor([0, 1, 2, 3, 4, 5]), (3, 2)) (tensor([0, 0, 1, 1, 2, 2]), tensor([0, 1, 0, 1, 0, 1])) >>> torch.unravel_index(torch.tensor([1234, 5678]), (10, 10, 10, 10)) (tensor([1, 5]), tensor([2, 6]), tensor([3, 7]), tensor([4, 8])) >>> torch.unravel_index(torch.tensor([[1234], [5678]]), (10, 10, 10, 10)) (tensor([[1], [5]]), tensor([[2], [6]]), tensor([[3], [7]]), tensor([[4], [8]])) >>> torch.unravel_index(torch.tensor([[1234], [5678]]), (100, 100)) (tensor([[12], [56]]), tensor([[34], [78]])) rFr1)rrr'_unravel_indexunbind)rrF res_tensors r+r'r'~s<h (($]WJUU/J   R  r,c^^^[R"TR5(+=(a: TR5(+=(a TR[R :X+U4Sj5 [R"[ T[[R[45U4Sj5 [ T[[R45(a[R"T/5mOXTH<m[R"[ T[[R45U4Sj5 M> [R"T5m[R"[ST55U4Sj5 [[[[R "[TSS[R"S/5-5["R$S9555nTR'S5R)[R*"UTR,[R.S 95[R*"TTR,[R.S 9-$) Nc">STR3$)Nz0expected 'indices' to be integer dtype, but got r)rsr+ _unravel_index..sB7==/Rr,c >S[T53$)Nz8expected 'shape' to be int or sequence of ints, but got rrsr+rr sJ4PU;-Xr,c >S[T53$)Nz8expected 'shape' sequence to only contain ints, but got r rRsr+rr sRSWX[S\R]^r,c3*# UH oS:v M g7f)rNr3)rhrOs r+ri!_unravel_index..s &1Hsc >S[T53$)Nz-'shape' cannot have negative values, but got )r<rsr+rr s?e ~Nr,r.)funcr1)r5r)r6 _check_type is_complexis_floating_pointrboolr9r:r;rrB _check_valueallr=reversedr{ accumulateoperatormul unsqueeze floor_dividerMr5int64)rrFcoefsrOs`` @r+rrs       ,))+ + ,  +R  53 h78X %#u||,-- E7#C   3ell 34^   5!  & &&N  $$U12YQC89    E   R - - U7>>E  U7>>E FFr,rc[U5(a[[U/UQ76$Uc[R"U5$[R"XS9$)aReturns the matrix product of the :math:`N` 2-D tensors. This product is efficiently computed using the matrix chain order algorithm which selects the order in which incurs the lowest cost in terms of arithmetic operations (`[CLRS]`_). Note that since this is a function to compute the product, :math:`N` needs to be greater than or equal to 2; if equal to 2 then a trivial matrix-matrix product is returned. If :math:`N` is 1, then this is a no-op - the original matrix is returned as is. .. warning:: :func:`torch.chain_matmul` is deprecated and will be removed in a future PyTorch release. Use :func:`torch.linalg.multi_dot` instead, which accepts a list of two or more tensors rather than multiple arguments. Args: matrices (Tensors...): a sequence of 2 or more 2-D tensors whose product is to be determined. out (Tensor, optional): the output tensor. Ignored if :attr:`out` = ``None``. Returns: Tensor: if the :math:`i^{th}` tensor was of dimensions :math:`p_{i} \times p_{i + 1}`, then the product would be of dimensions :math:`p_{1} \times p_{N + 1}`. Example:: >>> # xdoctest: +SKIP >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> a = torch.randn(3, 4) >>> b = torch.randn(4, 5) >>> c = torch.randn(5, 6) >>> d = torch.randn(6, 7) >>> # will raise a deprecation warning >>> torch.chain_matmul(a, b, c, d) tensor([[ -2.3375, -3.9790, -4.1119, -6.6577, 9.5609, -11.5095, -3.2614], [ 21.4038, 3.3378, -8.4982, -5.2457, -10.2561, -2.4684, 2.7163], [ -0.9647, -5.8917, -2.3213, -5.2284, 12.8615, -12.2816, -2.5095]]) .. _`[CLRS]`: https://mitpress.mit.edu/books/introduction-algorithms-third-edition r)rrrr)rmatricess r+rrsIL(##$\8GhGG {))22r,c6[R"XU(+S9$)afComputes the LU factorization of a matrix or batches of matrices :attr:`A`. Returns a tuple containing the LU factorization and pivots of :attr:`A`. Pivoting is done if :attr:`pivot` is set to ``True``. .. warning:: :func:`torch.lu` is deprecated in favor of :func:`torch.linalg.lu_factor` and :func:`torch.linalg.lu_factor_ex`. :func:`torch.lu` will be removed in a future PyTorch release. ``LU, pivots, info = torch.lu(A, compute_pivots)`` should be replaced with .. code:: python LU, pivots = torch.linalg.lu_factor(A, compute_pivots) ``LU, pivots, info = torch.lu(A, compute_pivots, get_infos=True)`` should be replaced with .. code:: python LU, pivots, info = torch.linalg.lu_factor_ex(A, compute_pivots) .. note:: * The returned permutation matrix for every matrix in the batch is represented by a 1-indexed vector of size ``min(A.shape[-2], A.shape[-1])``. ``pivots[i] == j`` represents that in the ``i``-th step of the algorithm, the ``i``-th row was permuted with the ``j-1``-th row. * LU factorization with :attr:`pivot` = ``False`` is not available for CPU, and attempting to do so will throw an error. However, LU factorization with :attr:`pivot` = ``False`` is available for CUDA. * This function does not check if the factorization was successful or not if :attr:`get_infos` is ``True`` since the status of the factorization is present in the third element of the return tuple. * In the case of batches of square matrices with size less or equal to 32 on a CUDA device, the LU factorization is repeated for singular matrices due to the bug in the MAGMA library (see magma issue 13). * ``L``, ``U``, and ``P`` can be derived using :func:`torch.lu_unpack`. .. warning:: The gradients of this function will only be finite when :attr:`A` is full rank. This is because the LU decomposition is just differentiable at full rank matrices. Furthermore, if :attr:`A` is close to not being full rank, the gradient will be numerically unstable as it depends on the computation of :math:`L^{-1}` and :math:`U^{-1}`. Args: A (Tensor): the tensor to factor of size :math:`(*, m, n)` pivot (bool, optional): controls whether pivoting is done. Default: ``True`` get_infos (bool, optional): if set to ``True``, returns an info IntTensor. Default: ``False`` out (tuple, optional): optional output tuple. If :attr:`get_infos` is ``True``, then the elements in the tuple are Tensor, IntTensor, and IntTensor. If :attr:`get_infos` is ``False``, then the elements in the tuple are Tensor, IntTensor. Default: ``None`` Returns: (Tensor, IntTensor, IntTensor (optional)): A tuple of tensors containing - **factorization** (*Tensor*): the factorization of size :math:`(*, m, n)` - **pivots** (*IntTensor*): the pivots of size :math:`(*, \text{min}(m, n))`. ``pivots`` stores all the intermediate transpositions of rows. The final permutation ``perm`` could be reconstructed by applying ``swap(perm[i], perm[pivots[i] - 1])`` for ``i = 0, ..., pivots.size(-1) - 1``, where ``perm`` is initially the identity permutation of :math:`m` elements (essentially this is what :func:`torch.lu_unpack` is doing). - **infos** (*IntTensor*, *optional*): if :attr:`get_infos` is ``True``, this is a tensor of size :math:`(*)` where non-zero values indicate whether factorization for the matrix or each minibatch has succeeded or failed Example:: >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_LAPACK) >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> A = torch.randn(2, 3, 3) >>> A_LU, pivots = torch.lu(A) >>> A_LU tensor([[[ 1.3506, 2.5558, -0.0816], [ 0.1684, 1.1551, 0.1940], [ 0.1193, 0.6189, -0.5497]], [[ 0.4526, 1.2526, -0.3285], [-0.7988, 0.7175, -0.9701], [ 0.2634, -0.9255, -0.3459]]]) >>> pivots tensor([[ 3, 3, 3], [ 3, 3, 3]], dtype=torch.int32) >>> A_LU, pivots, info = torch.lu(A, get_infos=True) >>> if info.nonzero().size(0) == 0: ... print('LU factorization succeeded for all samples!') LU factorization succeeded for all samples! )pivot check_errors)r6 _lu_with_info)rYr" get_infosrs r+_lu_implr&sB   q] LLr,out_lenr%cU(aSOSnX- S:wa[SS[U5-SU35e[U[[45(d![S[ U5R 35eg)Nr.rrUzexpected tuple of z elements but got z-argument 'out' must be tuple of Tensors, not ) TypeErrorr:r9r<r=r__name__)r'r%r get_infos_ints r+_check_list_sizer,{sx"AM!# S^!3 44Fwi P   cE4= ) );DIr@ resize_as_copy_rYr"r%rrJrKs r+_lu_with_infosr3s""$ a   a /F S92s3xA F  fi ( . .vy 9!  r,c &[U5(a[[U4XX#S9$[XX#5nUbX[ [ U5X#5 [ [ U55H)nX5RXE5RXE5 M+ U$USUS4$)Nr.rr.r/r2s r+ _lu_no_infosr5s ""$ a   a /F S92s3xA F  fi ( . .vy 9! ay&)##r,rc[S5e)Nz$`align_tensors` not yet implemented.)rAr)s r+rrs = >>r,)r) NNNTreflectFNNN)TFFN)FFN)rUNrf)rr)rNFNN)TFN)Qr{rcollections.abcrtypingrrrrr6torch.nn.functionalnn functionalrrr torch._Cr torch._jit_internalr overloadr torch._lowrankr rtorch.overridesrrrr__all__rrr:r=r<r"rrur!rrr#r_unique_impl_outrrrrrr*_return_inverse_false_return_inverse_truer%__doc__rrr!_consecutive_return_inverse_false _consecutive_return_inverse_truer&r$rrrrrrr floatrBr'rrr& _ListOrSeqr,r3r5rrr3r,r+rKs"$66  G3 6*@W$z45 45!#tCy.145 45 63; 45nO5#O5&O5fJN6V ,-69A#6 vs{ 66:^6Xc]^6eFCK>P^6B +(3- +$!% $##%)&*w w w w w V  w  wwwtnwTNwd^w wt  KKMP fVVV34   f+ f+ f+f+ f+ # f+  f+V! C+ C+C+C+ # C+  C+P  $  $  #$)    (        " %%  & " #"%5   & '"%!$4   $ ("$ &   , ."6== &*   ell #    '+  DItCy() ell #    '+  49o ell #    '+  ll ell #    "& c< %,,  c