JTh/DSrSSKrSSKrSSKrSSKrSSKJrJrJ r SSK r \ R\ R\ R\ R\ R\ R \ R"\ R$/r\ R(\ R*\ R,\ R./r\ R2\ R4\ R6\ R8/r\ R<\ R>\ R@/r!\ RD/\Qr#/\Q\!Qr$S\ RJS\&S\&S\ RJ4Sjr'SSS S S SS .S \ \(\ RR\*\(\+\(S 44S \ RXS\ \-\ R\4S\\&S\\&S\"S\"S\"S\\ R^S\ RJ4Sjjr0g)z1 This module contains tensor creation utilities. N)castOptionalUniontlowhighreturncX!- [R"UR5R:a'UR US- US- 5R S5$UR X5$)N)torchfinfodtypemaxuniform_mul_)rrrs O/var/www/auris/envauris/lib/python3.13/site-packages/torch/testing/_creation.py_uniform_random_r$sT zU[[)---zz#'4!8,11!44zz#$$F)rr requires_grad noncontiguous exclude_zero memory_formatshape.rdevicerrrrc^S[[S[[S[S[S[S[S[[[44U4Sjjn [U5S :Xa1[ US [ R R5(aUS n[[[S 4[U55nU(aUb[S U<SU<35eU(aT[;a[ST<35eU=(a [R"SUS 5S :nU(a'[[[S 4/US SQSUS-P75nT[RLa@[[[[4U "UUS SS SS95up#[R "X#XTS9n GOT[;a[[[[4U "UU[R""T5R$[R""T5R&T[R(LaS OS -SSS95up#[R "X#XTS9n GO8T[*;aU "UU[R,"T5R$[R,"T5R&SSS9up#[R."XTS9n [1T[2;a[R4"U 5OU X#5 OT[6;aU "UU[R,"T5R$[R,"T5R&SSS9up#[R."X[R8S9n [1XU5 U R;T5n O[=STS35eU(a U S S S S24n OUbU R?US9n U(a1T[;aS O[R,"T5R@XS :H'T[*;aXJl!U $)aCreates a tensor with the given :attr:`shape`, :attr:`device`, and :attr:`dtype`, and filled with values uniformly drawn from ``[low, high)``. If :attr:`low` or :attr:`high` are specified and are outside the range of the :attr:`dtype`'s representable finite values then they are clamped to the lowest or highest representable finite value, respectively. If ``None``, then the following table describes the default values for :attr:`low` and :attr:`high`, which depend on :attr:`dtype`. +---------------------------+------------+----------+ | ``dtype`` | ``low`` | ``high`` | +===========================+============+==========+ | boolean type | ``0`` | ``2`` | +---------------------------+------------+----------+ | unsigned integral type | ``0`` | ``10`` | +---------------------------+------------+----------+ | signed integral types | ``-9`` | ``10`` | +---------------------------+------------+----------+ | floating types | ``-9`` | ``9`` | +---------------------------+------------+----------+ | complex types | ``-9`` | ``9`` | +---------------------------+------------+----------+ Args: shape (Tuple[int, ...]): Single integer or a sequence of integers defining the shape of the output tensor. dtype (:class:`torch.dtype`): The data type of the returned tensor. device (Union[str, torch.device]): The device of the returned tensor. low (Optional[Number]): Sets the lower limit (inclusive) of the given range. If a number is provided it is clamped to the least representable finite value of the given dtype. When ``None`` (default), this value is determined based on the :attr:`dtype` (see the table above). Default: ``None``. high (Optional[Number]): Sets the upper limit (exclusive) of the given range. If a number is provided it is clamped to the greatest representable finite value of the given dtype. When ``None`` (default) this value is determined based on the :attr:`dtype` (see the table above). Default: ``None``. .. deprecated:: 2.1 Passing ``low==high`` to :func:`~torch.testing.make_tensor` for floating or complex types is deprecated since 2.1 and will be removed in 2.3. Use :func:`torch.full` instead. requires_grad (Optional[bool]): If autograd should record operations on the returned tensor. Default: ``False``. noncontiguous (Optional[bool]): If `True`, the returned tensor will be noncontiguous. This argument is ignored if the constructed tensor has fewer than two elements. Mutually exclusive with ``memory_format``. exclude_zero (Optional[bool]): If ``True`` then zeros are replaced with the dtype's small positive value depending on the :attr:`dtype`. For bool and integer types zero is replaced with one. For floating point types it is replaced with the dtype's smallest positive normal number (the "tiny" value of the :attr:`dtype`'s :func:`~torch.finfo` object), and for complex types it is replaced with a complex number whose real and imaginary parts are both the smallest positive normal number representable by the complex type. Default ``False``. memory_format (Optional[torch.memory_format]): The memory format of the returned tensor. Mutually exclusive with ``noncontiguous``. Raises: ValueError: If ``requires_grad=True`` is passed for integral `dtype` ValueError: If ``low >= high``. ValueError: If either :attr:`low` or :attr:`high` is ``nan``. ValueError: If both :attr:`noncontiguous` and :attr:`memory_format` are passed. TypeError: If :attr:`dtype` isn't supported by this function. Examples: >>> # xdoctest: +SKIP >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA) >>> from torch.testing import make_tensor >>> # Creates a float tensor with values in [-1, 1) >>> make_tensor((3,), device='cpu', dtype=torch.float32, low=-1, high=1) >>> # xdoctest: +SKIP tensor([ 0.1205, 0.2282, -0.6380]) >>> # Creates a bool tensor on CUDA >>> make_tensor((2, 2), device='cuda', dtype=torch.bool) tensor([[False, False], [False, True]], device='cuda:0') rrlowest_inclusivehighest_exclusive default_low default_highr c >S[S[S[S[4SjnUbUOUnUbUOUn[SX455(a[SU<SU<35eX:Xa%T[;a[R "S [ S S 9 O;X:a[S US U35eX:dX:a[SUSUSTSUSUS3 5eU"XU5nU"XU5nT[;a,[R"U5[R"U54$X4$)z~ Modifies (and raises ValueError when appropriate) low and high values given by the user (input_low, input_high) if required. alhr c,[[X5U5$N)minr)r!r"r#s rclamp3make_tensor..modify_low_high..clampss1y!$ $rc3|# UH2n[U[5=(a [R"U5v M4 g7fr%) isinstancefloatmathisnan).0values r 7make_tensor..modify_low_high..s(W;%z%'=DJJu,==;s:<z,`low` and `high` cannot be NaN, but got low=z and high=zPassing `low==high` to `torch.testing.make_tensor` for floating or complex types is deprecated since 2.1 and will be removed in 2.3. Use `torch.full(...)` instead.) stacklevelz(`low` must be less than `high`, but got z >= z5The value interval specified by `low` and `high` is [z, z), but z only supports [)) r+any ValueError_FLOATING_OR_COMPLEX_TYPESwarningswarn FutureWarning_BOOLEAN_OR_INTEGRAL_TYPESr,ceil)rrrrrrr'rs rmodify_low_high$make_tensor..modify_low_highsK %U %u % %5 %_c+'t\ WC;W W W?3& dWM [U&@@ MM1  [GuDQUPVWX X  $(@GuBtfUg-.>-?rBSATTUW  C+<=T->? . .99S>499T?2 2yrr.NzaThe parameters `noncontiguous` and `memory_format` are mutually exclusive, but got noncontiguous=z and memory_format=zU`requires_grad=True` is not supported for boolean and integral dtypes, but got dtype=c X-$r%)xys rmake_tensor..sAErr )rrrr)rri zThe requested dtype 'z' is not supported by torch.testing.make_tensor(). To request support, file an issue at: https://github.com/pytorch/pytorch/issues)r)"rr+tuplelenr* collectionsabcSequencerintr6r; functoolsreducer boolrandintiinfor&rint64r7r emptyr_COMPLEX_TYPES view_as_real_FLOATING_8BIT_TYPESfloat32to TypeErrorclonetinyr) rrrrrrrrrr=results ` r make_tensorr_-sd1 e_1uo1 1 ! 1  11 ue| 1f 5zQ:eAh 0H0HIIa sCx%, /E2&$&&:M+; =  "<<d^c]e f  "Xi&6&67I5RS&TWX&XMU38_&Bcr &BAb M&BC   #s(O !""#    s%eL , , #s(O !&U!3!7!7"'++e"4"8"8U[[01a #9  "s%eL , ,#  "[[/33#kk%044  U?*/>*AE  v &vs  & &#  "[[/33#kk%044  UGd+5!#E7+_ _  QTT "  "M:44A%++e:L:Q:Q { **, Mr)1__doc__collections.abcrKrOr,r8typingrrrr uint8int8int16int32rTuint16uint32uint64_INTEGRAL_TYPESfloat16bfloat16rYfloat64_FLOATING_TYPES float8_e4m3fn float8_e5m2float8_e4m3fnuzfloat8_e5m2fnuzrX complex32 complex64 complex128rVrQr;r7Tensorr+rrNSizelistrIrstrrrr_rArrrzs ((  KK JJ KK KK KK LL LL LL ==%..%--O      //5??E4D4DE#jj;?;@@@% %5%%%,,%  37g #uzz49eCHo= >g ;;g #u||# $g % g 5/ g  gggE//0g \\gr