sg)ddlZddlmZddlZddlmZddlmZmZddl m Z m Z ddl m Z ddlmZgd ZGd d eZGd d eZGddeZGddeZGdde eZy)N)Any)Tensor) functionalinit) ParameterUninitializedParameter)LazyModuleMixin)Module)BilinearIdentity LazyLinearLinearc@eZdZdZdededdffd ZdedefdZxZS) r aA placeholder identity operator that is argument-insensitive. Args: args: any argument (unused) kwargs: any keyword argument (unused) Shape: - Input: :math:`(*)`, where :math:`*` means any number of dimensions. - Output: :math:`(*)`, same shape as the input. Examples:: >>> m = nn.Identity(54, unused_argument1=0.1, unused_argument2=False) >>> input = torch.randn(128, 20) >>> output = m(input) >>> print(output.size()) torch.Size([128, 20]) argskwargsreturnNc"t|yNsuper__init__)selfrr __class__s J/var/www/html/venv/lib/python3.12/site-packages/torch/nn/modules/linear.pyrzIdentity.__init__+s inputc|Srrrs rforwardzIdentity.forward.s r) __name__ __module__ __qualname____doc__rrrr! __classcell__rs@rr r s5(cSTVrr c eZdZUdZddgZeed<eed<eed< d dedededdffd Z dd Z d edefd Z de fd Z xZS)ra/Applies an affine linear transformation to the incoming data: :math:`y = xA^T + b`. This module supports :ref:`TensorFloat32`. On certain ROCm devices, when using float16 inputs this module will use :ref:`different precision` for backward. Args: in_features: size of each input sample out_features: size of each output sample bias: If set to ``False``, the layer will not learn an additive bias. Default: ``True`` Shape: - Input: :math:`(*, H_{in})` where :math:`*` means any number of dimensions including none and :math:`H_{in} = \text{in\_features}`. - Output: :math:`(*, H_{out})` where all but the last dimension are the same shape as the input and :math:`H_{out} = \text{out\_features}`. Attributes: weight: the learnable weights of the module of shape :math:`(\text{out\_features}, \text{in\_features})`. The values are initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})`, where :math:`k = \frac{1}{\text{in\_features}}` bias: the learnable bias of the module of shape :math:`(\text{out\_features})`. If :attr:`bias` is ``True``, the values are initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})` where :math:`k = \frac{1}{\text{in\_features}}` Examples:: >>> m = nn.Linear(20, 30) >>> input = torch.randn(128, 20) >>> output = m(input) >>> print(output.size()) torch.Size([128, 30]) in_features out_featuresweightNbiasrc&||d}t|||_||_t t j ||ffi||_|r%t t j |fi||_n|jdd|jyNdevicedtyper,) rrr)r*rtorchemptyr+r,register_parameterreset_parameters)rr)r*r,r0r1factory_kwargsrs rrzLinear.__init__]s%+U; &( KK{3 F~ F   !%++l"Mn"MNDI  # #FD 1 rcLtj|jtjd|j dtj |j\}}|dkDrdtj|z nd}tj|j | |yy)N)arr )rkaiming_uniform_r+mathsqrtr,_calculate_fan_in_and_fan_outuniform_)rfan_in_bounds rr5zLinear.reset_parametersrsx dkkTYYq\: 99 ::4;;GIFA-3aZA &))QE MM$))eVU 3 !rrcXtj||j|jSr)Flinearr+r,r s rr!zLinear.forward|sxxt{{DII66rcXd|jd|jd|jduS)Nz in_features=, out_features=, bias=)r)r*r,rs r extra_reprzLinear.extra_reprs8d../t?P?P>QQXY]YbYbjnYnXopprTNNrNr"r#r$r% __constants__int__annotations__rboolrr5r!strrIr&r's@rrr2s#J#N3M N         *47V77qCqrrc 8eZdZ ddedededdffd ZxZS)NonDynamicallyQuantizableLinearNr)r*r,rc.t||||||y)N)r,r0r1r)rr)r*r,r0r1rs rrz(NonDynamicallyQuantizableLinear.__init__s"  Du  rrJ)r"r#r$rNrPrr&r's@rrSrSs>             rrSc eZdZUdZgdZeed<eed<eed<eed< ddedededed df fd Z dd Z d ed ed efdZ d e fdZ xZS)r aApplies a bilinear transformation to the incoming data: :math:`y = x_1^T A x_2 + b`. Args: in1_features: size of each first input sample in2_features: size of each second input sample out_features: size of each output sample bias: If set to False, the layer will not learn an additive bias. Default: ``True`` Shape: - Input1: :math:`(*, H_{in1})` where :math:`H_{in1}=\text{in1\_features}` and :math:`*` means any number of additional dimensions including none. All but the last dimension of the inputs should be the same. - Input2: :math:`(*, H_{in2})` where :math:`H_{in2}=\text{in2\_features}`. - Output: :math:`(*, H_{out})` where :math:`H_{out}=\text{out\_features}` and all but the last dimension are the same shape as the input. Attributes: weight: the learnable weights of the module of shape :math:`(\text{out\_features}, \text{in1\_features}, \text{in2\_features})`. The values are initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})`, where :math:`k = \frac{1}{\text{in1\_features}}` bias: the learnable bias of the module of shape :math:`(\text{out\_features})`. If :attr:`bias` is ``True``, the values are initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})`, where :math:`k = \frac{1}{\text{in1\_features}}` Examples:: >>> m = nn.Bilinear(20, 30, 40) >>> input1 = torch.randn(128, 20) >>> input2 = torch.randn(128, 30) >>> output = m(input1, input2) >>> print(output.size()) torch.Size([128, 40]) ) in1_features in2_featuresr*rVrWr*r+Nr,rc6||d}t|||_||_||_t t j|||ffi||_|r%t t j|fi||_ n|jdd|jyr.) rrrVrWr*rr2r3r+r,r4r5) rrVrWr*r,r0r1r6rs rrzBilinear.__init__s%+U; ((( KK|\B Un U   !%++l"Mn"MNDI  # #FD 1 rcdtj|jjdz }t j |j| ||j #t j |j | |yy)Nr )r;r<r+sizerr>r,)rrAs rr5zBilinear.reset_parameterss_DIIdkk..q122 dkkE651 99 MM$))eVU 3 !rinput1input2cZtj|||j|jSr)rCbilinearr+r,)rr[r\s rr!zBilinear.forwardszz&&$++tyyAArc rd|jd|jd|jd|jduS)Nz in1_features=z, in2_features=rFrG)rVrWr*r,rHs rrIzBilinear.extra_reprsJD--.od>O>O=PQ --.gdiit6K5L N rrJrKrLr's@rr r s#JEM N         04 BfBfBB C rr cbeZdZUdZeZeed<eed< d dede ddffd Z d fd Z d d Z xZ S) raA :class:`torch.nn.Linear` module where `in_features` is inferred. In this module, the `weight` and `bias` are of :class:`torch.nn.UninitializedParameter` class. They will be initialized after the first call to ``forward`` is done and the module will become a regular :class:`torch.nn.Linear` module. The ``in_features`` argument of the :class:`Linear` is inferred from the ``input.shape[-1]``. Check the :class:`torch.nn.modules.lazy.LazyModuleMixin` for further documentation on lazy modules and their limitations. Args: out_features: size of each output sample bias: If set to ``False``, the layer will not learn an additive bias. Default: ``True`` Attributes: weight: the learnable weights of the module of shape :math:`(\text{out\_features}, \text{in\_features})`. The values are initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})`, where :math:`k = \frac{1}{\text{in\_features}}` bias: the learnable bias of the module of shape :math:`(\text{out\_features})`. If :attr:`bias` is ``True``, the values are initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})` where :math:`k = \frac{1}{\text{in\_features}}` r+r,Nr*rc||d}t|dddtdi||_||_|rtdi||_yy)Nr/rFr)rrrr+r*r,)rr*r,r0r1r6rs rrzLazyLinear.__init__ sP%+U; Au%,>~> ( .@@DI rcd|js|jdk7rt| yyy)Nr)has_uninitialized_paramsr)rr5)rrs rr5zLazyLinear.reset_parameterss0,,.43C3Cq3H G $ &4I.rc|jrtj5|jd|_|j j |j|jf|j&|jj |jf|jdddyy#1swYyxYw)N) rcr2no_gradshaper)r+ materializer*r,r5r s rinitialize_parametersz LazyLinear.initialize_parameterss  ( ( * (#(;;r?  ''):):D?%%'  ( ( + ( (s BB77CrJrK)r"r#r$r%r cls_to_becomerrOrNrPrr5rir&r's@rrrsM8M ""  HL A A'+ A  A'(rr)r;typingrr2rtorch.nnrrCrtorch.nn.parameterrrlazyr moduler __all__r rrSr rrrrrqsj  *@! v8NqVNql  f  Q vQ h8(&8(r