tensorplay.nn.modules

The tensorplay.nn module provides the building blocks for creating neural networks. It is designed to be highly modular and compatible with the PyTorch nn API.

Core Concepts

• Module: The base class for all neural network layers. Your models should also inherit from this class.
• Parameter: A special kind of Tensor that is automatically registered as a module parameter when assigned as an attribute.
• Functional: Stateless versions of the layers, found in tensorplay.nn.functional.

Module Categories

For specific layers, see:

• Modules (Base): Base classes like Module and Parameter.
• Linear Layers: Fully connected layers.
• Convolution Layers: 1D, 2D, and 3D convolution layers.
• Pooling Layers: Max and Average pooling.
• Activation Functions: ReLU, Sigmoid, Tanh, etc.
• Normalization Layers: BatchNorm, LayerNorm, InstanceNorm.
• Loss Functions: MSE, CrossEntropy, NLLLoss.
• Dropout Layers: Dropout and AlphaDropout.
• Container: Sequential, ModuleList, ModuleDict.