o eiF @sDdZddlZddlmZddlmZeeZGdddejjZ dS)zbLibrary implementing linear transformation. Authors * Mirco Ravanelli 2020 * Davide Borra 2021 N) get_loggercs4eZdZdZ     d fdd ZddZZS) LinearaComputes a linear transformation y = wx + b. Arguments --------- n_neurons : int It is the number of output neurons (i.e, the dimensionality of the output). input_shape : tuple It is the shape of the input tensor. input_size : int Size of the input tensor. bias : bool If True, the additive bias b is adopted. max_norm : float weight max-norm. combine_dims : bool If True and the input is 4D, combine 3rd and 4th dimensions of input. Example ------- >>> inputs = torch.rand(10, 50, 40) >>> lin_t = Linear(input_shape=(10, 50, 40), n_neurons=100) >>> output = lin_t(inputs) >>> output.shape torch.Size([10, 50, 100]) NTFcsvt||_||_|dur|durtd|dur0|d}t|dkr0|jr0|d|d}tj|||d|_dS)Nz)Expected one of input_shape or input_size)bias) super__init__max_norm combine_dims ValueErrorlennnrw)self n_neurons input_shape input_sizerr r  __class__U/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/speechbrain/nnet/linear.pyr ,s zLinear.__init__cCsr|jdkr|jr||jd|jd|jd|jd}|jdur2tj|jjj dd|jd|jj_ ||}|S)a Returns the linear transformation of input tensor. Arguments --------- x : torch.Tensor Input to transform linearly. Returns ------- wx : torch.Tensor The linearly transformed outputs. rrrrN)pdimmaxnorm) ndimr reshapeshaper torchrenormrweightdata)rxwxrrrforwardDs *  zLinear.forward)NNTNF)__name__ __module__ __qualname____doc__r r& __classcell__rrrrrsr) r*r torch.nnrspeechbrain.utils.loggerrr'loggerModulerrrrrs