Conv2d

Conv2d is a 2-D convolutional layer that learns a bank of spatial filters applied to each input channel. It stores a weight tensor of shape (out_channels, in_channels, kH, kW) and an optional bias vector, both initialised with Kaiming uniform initialisation. The layer wraps the functional conv2d op and is fully compatible with Sequential and any Optimizer.

import simplegrad as sg
import simplegrad.nn as nn

conv = nn.Conv2d(in_channels=3, out_channels=16, kernel_size=3, padding=1)
x = sg.normal((1, 3, 32, 32))
out = conv(x)  # shape: (1, 16, 32, 32)

Conv2d

Bases: Module

2D convolutional layer.

Applies a learned 2D convolution over an input signal. Weights are initialized with Kaiming uniform scaling.

Parameters:

• in_channels (int | None, default: None ) –

  Number of input channels.

• out_channels (int | None, default: None ) –

  Number of output channels (filters).

• kernel_size (int | tuple[int, int], default: None ) –

  Kernel size. Int or (kH, kW).

• weight (Tensor | None, default: None ) –

  Optional pre-built weight tensor of shape (out_channels, in_channels, kH, kW).

• bias (Tensor | None, default: None ) –

  Optional pre-built bias tensor of shape (out_channels,).

• use_bias (bool, default: True ) –

  Add a bias term. Defaults to True.

• dtype (str | None, default: None ) –

  Data type string. Defaults to "float32".

• stride (int, default: 1 ) –

  Convolution stride. Int or (sH, sW). Defaults to 1.

• pad_width (int | tuple[int, int, int, int], default: 0 ) –

  Padding. Int (all sides) or (top, bottom, left, right). Defaults to 0.

• pad_mode (str, default: 'constant' ) –

  Padding mode. Defaults to "constant".

• pad_value (int, default: 0 ) –

  Fill value for constant padding. Defaults to 0.

• weight_label (str, default: 'W' ) –

  Label for the weight tensor.

• bias_label (str, default: 'b' ) –

  Label for the bias tensor.

Attributes

Attribute	Type	Description
.weight	Tensor	Filter bank of shape (out_channels, in_channels, kH, kW). Learnable.
.bias	Tensor \| None	Bias vector of shape (out_channels,). None if use_bias=False.
.in_channels	int	Number of input channels.
.out_channels	int	Number of output channels (filters).
.kernel_size	int \| tuple	Height and width of the convolution kernel.
.stride	int \| tuple	Step size of the sliding window.
.padding	int \| tuple	Zero-padding added to both spatial dimensions.

Methods

Method	Description
.forward()	Apply the convolution to input tensor x.

Inherits all methods from Module: .parameters(), .submodules(), .to_device(), .summary(), .set_train_mode(), .set_eval_mode().