.. currentmodule:: xlnstorch.nn .. _layers-doc: Layers ====== The ``xlnstorch`` package provides a set of layers that support ``LNSTensor`` parameters. Whilst you can use the built-in PyTorch layers with ``LNSTensor`` inputs, these layers will implement standard floating-point parameters. Note that some of the layers in this module are only added for completeness since they don't have any parameters. For example, the ``nn.LNSDropout`` layer is equivalent to the standard PyTorch ``torch.nn.Dropout`` layer, both will work with ``LNSTensor`` inputs, but the ``nn.LNSDropout`` layer is implemented for completeness. Any layers that do not have parameters will be denoted in the documentation. =============================== ===================== Linear Layers Note =============================== ===================== :class:`LNSIdentity` No parameters :class:`LNSLinear` :class:`LNSBilinear` :class:`LNSLazyLinear` =============================== ===================== =============================== ===================== Dropout Layers Note =============================== ===================== :class:`LNSDropout` No parameters :class:`LNSDropout1d` No parameters :class:`LNSDropout2d` No parameters :class:`LNSDropout3d` No parameters =============================== ===================== =============================== ===================== Convolutional Layers Note =============================== ===================== :class:`LNSConv1d` :class:`LNSConv2d` :class:`LNSConv3d` =============================== ===================== =============================== ===================== Normalization Layers Note =============================== ===================== :class:`LNSBatchNorm1d` :class:`LNSBatchNorm2d` :class:`LNSBatchNorm3d` :class:`LNSLayerNorm` =============================== ===================== =============================== ===================== Pooling Layers Note =============================== ===================== :class:`LNSAvgPool1d` No parameters :class:`LNSAvgPool2d` No parameters :class:`LNSAvgPool3d` No parameters :class:`LNSAdaptiveAvgPool1d` No parameters :class:`LNSAdaptiveAvgPool2d` No parameters :class:`LNSAdaptiveAvgPool3d` No parameters :class:`LNSMaxPool1d` No parameters :class:`LNSMaxPool2d` No parameters :class:`LNSMaxPool3d` No parameters =============================== ===================== =============================== ===================== Recurrent Layers Note =============================== ===================== :class:`LNSRNN` :class:`LNSRNNCell` :class:`LNSLSTM` :class:`LNSLSTMCell` :class:`LNSGRU` :class:`LNSGRUCell` =============================== ===================== =============================== ===================== Transformer Layers Note =============================== ===================== :class:`LNSMultiheadAttention` =============================== ===================== .. hide the autosummary table from the main page but still .. generate the stub files for the layers (this is a hack). Custom Layers ------------- To implement your own custom layers that support ``LNSTensor`` parameters, you can subclass the base layer class provided in this module :class:`LNSModule`. This base class is a subclass of the standard PyTorch ``torch.nn.Module`` and provides the method :func:`LNSModule.register_parameter` which is equivalent to PyTorch's method of registering parameters. .. raw:: html
.. autosummary:: :toctree: generated/nn :nosignatures: LNSModule LNSIdentity LNSLinear LNSBilinear LNSLazyLinear LNSDropout LNSDropout1d LNSDropout2d LNSDropout3d LNSConv1d LNSConv2d LNSConv3d LNSBatchNorm1d LNSBatchNorm2d LNSBatchNorm3d LNSLayerNorm LNSAvgPool1d LNSAvgPool2d LNSAvgPool3d LNSAdaptiveAvgPool1d LNSAdaptiveAvgPool2d LNSAdaptiveAvgPool3d LNSMaxPool1d LNSMaxPool2d LNSMaxPool3d LNSRNN LNSRNNCell LNSLSTM LNSLSTMCell LNSGRU LNSGRUCell LNSMultiheadAttention .. raw:: html