Expand description
Neural network layers and network types (FP32, INT8, INT4).
Structs§
- Conv2d
Layer - 2D convolution layer (FP32).
- Conv2d
LayerQ - INT8 quantized 2D convolution layer.
- Conv2d
Layer Q4 - INT4 quantized 2D convolution layer.
- Flatten
Layer - Flatten layer that reshapes 4D tensors into 2D (FP32).
- Flatten
LayerQ - INT8 quantized flatten layer.
- Flatten
Layer Q4 - INT4 quantized flatten layer.
- Linear
Layer - Fully-connected (linear) layer (FP32).
- Linear
LayerQ - INT8 quantized fully-connected (linear) layer.
- Linear
Layer Q4 - INT4 quantized fully-connected (linear) layer.
- MaxPool2d
Layer - 2D max pooling layer (FP32).
- MaxPool2d
LayerQ - INT8 quantized 2D max pooling layer.
- MaxPool2d
Layer Q4 - INT4 quantized 2D max pooling layer.
- Neural
Network - A FP32 neural network composed of sequential layers.
- Quantized
Neural Network - An INT8 quantized neural network.
- Quantized
Neural Network I4 - An INT4 quantized neural network.
- ReLu
Layer - ReLU activation layer (FP32).
- ReLu
LayerQ - INT8 quantized ReLU activation layer.
- ReLu
Layer Q4 - INT4 quantized ReLU activation layer.
- Soft
MaxLayer - Softmax activation layer (FP32).
Enums§
- Layer
Type - Identifies the type of a neural network layer.
Traits§
- Layer
- Trait for FP32 neural network layers.
- Quantized
Layer - Trait for INT8 quantized layers.
- Quantized
Layer I4 - Trait for INT4 quantized layers.