ёiM%*SSKJr SSKrSSKrSSKJr SSKJrJrJrJ r SSK r SSK J r SSK Jr SSKJr \(aSS KJr SS K Jr SS KJr S S /r\ "SSS9r"SS\\5r"SS\R5rSSjrSSS.SSjjrg)) annotationsN)IntEnum) TYPE_CHECKINGLiteralProtocolTypeVar) DenseTensor) check_type)in_dygraph_mode) CapsuleType)Tensor) PlaceLike to_dlpack from_dlpack _T_contraT) contravariantcL\rSrSrSrSSSSS.S SjjrS SjrSrg) SupportDLPack+z ref: https://github.com/numpy/numpy/blob/7e6e48ca7aacae9994d18a3dadbabd2b91c32151/numpy/__init__.pyi#L3068-L3077 https://github.com/numpy/numpy/blob/7e6e48ca7aacae9994d18a3dadbabd2b91c32151/numpy/__init__.pyi#L4730-L4731 .N)stream max_version dl_devicecopycgN)selfrrrrs S/var/www/html/banglarbhumi/venv/lib/python3.13/site-packages/paddle/utils/dlpack.py __dlpack__SupportDLPack.__dlpack__2scgrr)rs r__dlpack_device__SupportDLPack.__dlpack_device__;s3r"r) rzNone | _T_contrarztuple[int, int] | Nonerztuple[IntEnum, int] | Noner bool | Nonereturnr )r'ztuple[int, Literal[0]])__name__ __module__ __qualname____firstlineno____doc__r r$__static_attributes__rr"rrr+sQ$'.104 !,  .    ?r"rcT\rSrSrSrSrSrSrSrSr Sr S r S r S r S rS rSrSrSrSrSrg) DLDeviceType>))r ))))) ) ) ) ) )))))rN)r(r)r*r+kDLCPUkDLCUDA kDLCUDAHost kDLOpenCL kDLVulkankDLMetalkDLVPIkDLROCM kDLROCMHost kDLExtDevkDLCUDAManaged kDLOneAPI kDLWebGPU kDLHexagonkDLMAIAkDLTrnr-rr"rr/r/>sR FGKIIH FGKINIIJG Fr"r/c*[5(ac[U[R5(d[ S[ U5S35eUR 5R5R5$[US[S5 UR5$)a Encodes a tensor to DLPack. Args: x (Tensor): The input tensor, and the data type can be ``bool``, ``float16``, ``float32``, ``float64``, ``int8``, ``int16``, ``int32``, ``int64``, ``uint8``, ``complex64``, ``complex128``. Returns: dltensor, and the data type is PyCapsule. Examples: .. code-block:: python :name: code-paddle-to-paddle >>> import paddle >>> # x is a tensor with shape [2, 4] >>> x = paddle.to_tensor([[0.2, 0.3, 0.5, 0.9], ... [0.1, 0.2, 0.6, 0.7]]) >>> dlpack = paddle.to_dlpack(x) >>> print(dlpack) >>> # doctest: +SKIP('the address will change in every run') >>> #doctest: -SKIP >>> # dlpack capsule will be renamed to 'used_dltensor' after decoded >>> y = paddle.from_dlpack(dlpack) >>> print(dlpack) >>> # doctest: +SKIP('the address will change in every run') .. code-block:: python :name: code-paddle-to-torch >>> # doctest: +SKIP('torch will not be installed') >>> # type: ignore >>> # convert tensor from paddle to other framework using to_dlpack >>> import torch >>> x = paddle.randn([2, 4]).to(device="cpu") >>> y = torch.from_dlpack(paddle.to_dlpack(x)) >>> print(y.shape) torch.Size([2, 4]) >>> # doctest: -SKIP zAThe type of 'x' in to_dlpack must be paddle.Tensor, but received .xr) r isinstancepaddler TypeErrortypevalue get_tensor _to_dlpackr r )rRs rrrQs{\!V]]++!!%a ,  wwy##%0022q# k2 <<>r")devicerc[US5(GaD0nSUS'UbX#S'UbU[RRR U5n[RR R U5US'UR5nUS[R4;a[R"5 [R"S[S 9 [RRR!US 5nSSS5 US[R:HnU(aWR"S:XaS O WR"nXS 'UR$"S 0UD6n OUn [RR R+U 5n [-5(a"[R."XR15S 9n U $!,(df  N=f![&a& UR)S5 UR$"S 0UD6n Nf=f)a Decodes a DLPack to a tensor. The returned Paddle tensor will share the memory with the tensor from given dlpack. Args: dlpack (SupportDLPack | CapsuleType): A PyCapsule object with the dltensor, or that implements '__dlpack__' and '__dlpack_device__' methods. If `dlpack` is a tensor (or ndarray) object, it must support the `__dlpack__` protocol (i.e., have a `dlpack.__dlpack__` method). Otherwise `dlpack` may be a DLPack capsule, which is an opaque `PyCapsule` instance, typically produced by a `to_dlpack` function or method. device (PlaceLike, optional): The device of the returned tensor. If not specified, the device will be the same as that of the input `dlpack`. copy (bool, optional): Whether or not to copy the input. If True, the output tensor always copied. If False, the output tensor must never copied, and raise a BufferError in case a copy is deemed necessary. If None, the output tensor must reuse the existing memory buffer if possible and copy otherwise. Default: None. Returns: out (Tensor): A tensor decoded from DLPack. The data type of returned tensor can be one of: ``int32``, ``int64``, ``float16``, ``float32`` and ``float64``. The device of returned tensor can be one of: ``CPU``, ``CUDAPlace``, ``CUDAPinnedPlace``. Examples: .. code-block:: python :name: code-paddle-from-paddle >>> import paddle >>> # From DLPack capsule >>> x = paddle.to_tensor([[0.2, 0.3, 0.5, 0.9], ... [0.1, 0.2, 0.6, 0.7]], place="cpu") >>> dlpack = paddle.to_dlpack(x) >>> y = paddle.from_dlpack(dlpack) >>> # dlpack capsule will be renamed to 'used_dltensor' after decoded >>> print(dlpack) >>> # doctest: +SKIP('the address will change in every run') >>> # doctest: -SKIP >>> print(y) Tensor(shape=[2, 4], dtype=float32, place=Place(cpu), stop_gradient=True, [[0.20000000, 0.30000001, 0.50000000, 0.89999998], [0.10000000, 0.20000000, 0.60000002, 0.69999999]]) >>> # data of tensor x is shared with tensor y >>> y[0, 0] = 10.0 >>> print(x) Tensor(shape=[2, 4], dtype=float32, place=Place(gpu:0), stop_gradient=True, [[10. , 0.30000001, 0.50000000, 0.89999998], [0.10000000, 0.20000000, 0.60000002, 0.69999999]]) .. code-block:: python :name: code-paddle-from-numpy >>> # Directly from external tensor that implements '__dlpack__' and '__dlpack_device__' methods >>> import paddle >>> import numpy as np >>> x = np.array([[0.2, 0.3, 0.5, 0.9], ... [0.1, 0.2, 0.6, 0.7]]) >>> y = paddle.from_dlpack(x) >>> y[0, 0] = 10.0 >>> # data of tensor x is shared with tensor y >>> print(x) [[10. 0.3 0.5 0.9] [ 0.1 0.2 0.6 0.7]] r )r1r rNrrrignore)categoryr1r)placer)hasattrrTbase framework_get_paddle_placecoreplace_to_dl_devicer$r/rAwarningscatch_warningsfilterwarnings UserWarningrZcudacurrent_stream cuda_streamr rUpoprr r_place) dlpackrZrkwargsr^ dlpack_deviceris_gpu stream_ptrdlpack_outs rrrsZv|$$ &}  !6N  KK));;FCE"(++"2"2"E"Ee"LF; 002    4 46 6((*'';G++::=;KL+#1%)=)==F 2 2a 7V=O=O  *8  2''1&1G-3[[-=-=-I-I.CmmCzz|< J?+*  2 JJ} %''1&1G 2s6AF5G5 G-G65G6)rRrr'r )rnzSupportDLPack | CapsuleTyperZzPlaceLike | Nonerr&r'r) __future__renumrertypingrrrrrT base.corer base.data_feederr base.frameworkr typing_extensionsr rpaddle._typingr__all__rrr/rrrr"rr~s# << #),-(  Kt 4 ?HY'?&4<<&8| $ z 'z z  z  zr"