o %ݫi!@s`dZddlZddlmZmZmZmZmZmZm Z m Z ddl m Z e e ZGdddejjZdS)zuLibrary implementing quaternion-valued linear transformation. Authors * Titouan Parcollet 2020 * Drew Wagner 2024 N)QuaternionLinearCustomBackward affect_initcheck_quaternion_inputquaternion_initquaternion_linear_opquaternion_linear_rotation_op!renorm_quaternion_weights_inplace unitary_init) get_loggercs@eZdZdZ       d fdd Zejjd d ZZ S) QLineara This function implements a fully connected quaternion-valued linear layer: y = Wx + b. y, W, x and b are thus quaternion numbers. A quaternion number is written as: r + xi + yj + zk. A tensor of quaternion numbers x = [batch, 32] can be understood as [batch, 0:7] = R, [batch, 8:15] = Xi, [batch, 16:23] = Yi, and [batch, 24:31] = Xi. Thus the features dimension is cut in four (must be divisible by 4). Arguments --------- n_neurons : int It is the number of output neurons (i.e, the dimensionality of the output). Please note that these are quaternion-valued neurons. If 256 neurons are specified, the output dimension will be 1024. input_shape : tuple Expected size of the input. bias : bool If True, the additive bias b is adopted. init_criterion : str , optional (glorot, he). This parameter controls the initialization criterion of the weights. It is combined with weights_init to build the initialization method of the quaternion-valued weights (default "glorot"). weight_init : str, optional (quaternion, unitary). This parameter defines the initialization procedure of the quaternion-valued weights. "quaternion" will generate quaternion-valued weights following the init_criterion and the quaternion polar form. "unitary" will normalize the weights to lie on the unit circle (default "quaternion"). More details in: "Quaternion recurrent neural networks", Parcollet T. autograd : bool, optional When True, the default PyTorch autograd will be used. When False, a custom backpropagation will be used, reducing by a factor 3 to 4 the memory consumption. It is also 2x slower. This only works with spinor = False (default True). spinor : bool, optional When True, the layer will be turned into a spinor layer. More precisely W*x will be turned into W*x*W-1. The input x will be rotated by W such as in a spinor neural network. However, x MUST be a quaternion with the real part equal to zero. (0 + xi + yj + zk). Indeed, the rotation operation only acts on the vector part. Note that W will always be normalized before the rotation to ensure the quaternion algebra (default False). More details in: "Quaternion neural networks", Parcollet T. vector_scale : bool, optional The vector_scale is only used when spinor = True. In the context of a spinor neural network, multiple rotations of the input vector x are performed and summed. Hence, the norm of the output vector always increases with the number of layers, making the neural network instable with deep configurations. The vector_scale parameters are learnable parameters that acts like gates by multiplying the output vector with a small trainable parameter (default False). max_norm: float weight max-norm. Example ------- >>> inputs = torch.rand(10, 50, 40) >>> lin = QLinear(n_neurons=100, input_shape=inputs.shape, weight_init='unitary') >>> output = lin(inputs) >>> output.shape torch.Size([10, 50, 400]) Tglorot quaternionFNc st||_||_||_||_||_||_| |_t |t r#d|g}t ||dd|_ |j|_ tjt|j |j |_tjt|j |j |_tjt|j |j |_tjt|j |j |_|jrxtjjt|jjdd|_n t|jjd|_|jr|jrtjt|j |j |_tjj|jjn t|j |j d|_|rtjtd||_n td|d|_|jjdt t!d|j|_"t#|j|j|j|j|j"|dS)NF) requires_gradr)r unitary)$super__init__ n_neuronsinit_criterion weight_initautogradspinor vector_scalemax_norm isinstanceintr in_features out_featurestorchnn ParameterTensorr_weighti_weightj_weightk_weightzerosshape zero_kernelrequires_grad_ scale_paraminitxavier_uniform_databiasfill_rr winitr) selfr input_shaper0rrrrrr __class__a/home/ubuntu/.local/lib/python3.10/site-packages/speechbrain/nnet/quaternion_networks/q_linear.pyrYsr    zQLinear.__init__c Cs|jdurt|j|j|j|j|jd|jr=|jr-t||j|j|j|j|j |j |j }|St ||j|j|j|j|j }|S| }|dkrT|\}}}||||}t||j|j|j|j|j }|dkrq||||d}|S)zReturns the linear transformation of input tensor. Arguments --------- x : torch.Tensor Input to transform linearly. Returns ------- The linearly transformed input. N)rr)rrr$r%r&r'rrrr0r,r*rdimsizeviewrapply)r3xout input_dimbatchtimefear7r7r8forwardsZ & zQLinear.forward)Tr r TFFN) __name__ __module__ __qualname____doc__rr jitignorerD __classcell__r7r7r5r8r sCXr )rHr *speechbrain.nnet.quaternion_networks.q_opsrrrrrrrr speechbrain.utils.loggerr rEloggerr!Moduler r7r7r7r8s (