bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul

class bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul(dtype=torch.float32, *args, **kwargs)

A PyTorch module for binary matrix multiplication. This module wraps the BinaryMatMulFunction and includes parameters for the clipping values used in the binary operations. It is designed for integration into neural networks where binary operations can offer computational benefits.

dtype

Data type of the clipping parameters. Defaults to torch.float.

Type:

torch.dtype

x_clip

Clipping parameter for the first input tensor.

Type:

torch.nn.Parameter

y_clip

Clipping parameter for the second input tensor.

Type:

torch.nn.Parameter

Methods

__init__	Initializes the BinaryMatMul module with optional dtype argument for the clipping parameters.
forward	Forward pass of the BinaryMatMul module.
set_activation_scale	Sets the clipping parameters based on the input tensors.

Attributes

training

__init__(dtype=torch.float32, *args, **kwargs)

Initializes the BinaryMatMul module with optional dtype argument for the clipping parameters.

Parameters:

dtype (torch.dtype, optional) – The data type for the clipping parameters.

forward(x: Tensor, y: Tensor) → Tensor

Forward pass of the BinaryMatMul module. Ensures input tensors are of appropriate dimensions and applies the binary matrix multiplication function.

Parameters:

• x (torch.Tensor) – The first input tensor.

• y (torch.Tensor) – The second input tensor.

Returns:

The result of the binary matrix multiplication.

Return type:

torch.Tensor

set_activation_scale(x: Tensor, y: Tensor) → None

Sets the clipping parameters based on the input tensors. If the clipping values are not already set, they are initialized based on the mean absolute values of the inputs.

Parameters:

• x (torch.Tensor) – The first input tensor.

• y (torch.Tensor) – The second input tensor.