bitorch.layers.qconv3d.QConv3d_NoAct

class bitorch.layers.qconv3d.QConv3d_NoAct(*args: Any, weight_quantization: Optional[Union[Quantization, str]] = None, pad_value: Optional[float] = None, bias: bool = False, **kwargs: Any)

Methods

__init__	initialization function for padding and quantization.
forward	forward the input tensor through the quantized convolution layer.
reset_parameters	overwritten from _ConvNd to initialize weights

Attributes

__init__(*args: Any, weight_quantization: Optional[Union[Quantization, str]] = None, pad_value: Optional[float] = None, bias: bool = False, **kwargs: Any) → None

initialization function for padding and quantization.

Parameters:

• weight_quantization (Union[str, Quantization], optional) – quantization module or name of quantization function. Defaults to None.

• padding_value (float, optional) – value used for padding the input sequence. Defaults to None.

forward(input: Tensor) → Tensor

forward the input tensor through the quantized convolution layer.

Parameters:

input (Tensor) – input tensor

Returns:

the convoluted output tensor, computed with padded input and quantized weights.

Return type:

Tensor

reset_parameters() → None

overwritten from _ConvNd to initialize weights