Wrappers for Stateless Layers#

SpikingJelly 的 无状态层包装器 封装了 PyTorch 的标准层，并支持脉冲神经网络的时间步进模式。 欲获取更多信息，请查看 基本概念 。

SpikingJelly's stateless layer wrappers wrap PyTorch's standard layers and support the step mode of Spiking Neural Networks. See Basic Conception for more details.

class spikingjelly.activation_based.layer.stateless_wrapper.Conv1d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', step_mode='s')[源代码]#

基类：Conv1d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.Conv1d

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • in_channels (int)

  • out_channels (int)

  • kernel_size (tuple[int, ...])

  • stride (tuple[int, ...])

  • padding (str | tuple[int, ...])

  • dilation (tuple[int, ...])

  • groups (int)

  • bias (Tensor | None)

  • padding_mode (str)

Refer to torch.nn.Conv1d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', step_mode='s')[源代码]#

基类：Conv2d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.Conv2d

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • in_channels (int)

  • out_channels (int)

  • kernel_size (tuple[int, ...])

  • stride (tuple[int, ...])

  • padding (str | tuple[int, ...])

  • dilation (tuple[int, ...])

  • groups (int)

  • bias (Tensor | None)

  • padding_mode (str)

Refer to torch.nn.Conv2d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.Conv3d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', step_mode='s')[源代码]#

基类：Conv3d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.Conv3d

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • in_channels (int)

  • out_channels (int)

  • kernel_size (tuple[int, ...])

  • stride (tuple[int, ...])

  • padding (str | tuple[int, ...])

  • dilation (tuple[int, ...])

  • groups (int)

  • bias (Tensor | None)

  • padding_mode (str)

Refer to torch.nn.Conv3d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.Upsample(size=None, scale_factor=None, mode='nearest', align_corners=None, recompute_scale_factor=None, step_mode='s')[源代码]#

基类：Upsample, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • size (int | tuple[int, ...] | None)

  • scale_factor (float | tuple[float, ...] | None)

  • mode (str)

  • align_corners (bool | None)

  • recompute_scale_factor (bool | None)

其他的参数API参见 torch.nn.Upsample

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • size (int | tuple[int, ...] | None)

  • scale_factor (float | tuple[float, ...] | None)

  • mode (str)

  • align_corners (bool | None)

  • recompute_scale_factor (bool | None)

Refer to torch.nn.Upsample for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

返回类型:

Tensor

class spikingjelly.activation_based.layer.stateless_wrapper.ConvTranspose1d(in_channels, out_channels, kernel_size, stride=1, padding=0, output_padding=0, groups=1, bias=True, dilation=1, padding_mode='zeros', step_mode='s')[源代码]#

基类：ConvTranspose1d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.ConvTranspose1d

  • English

参数:

Refer to torch.nn.ConvTranspose1d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.ConvTranspose2d(in_channels, out_channels, kernel_size, stride=1, padding=0, output_padding=0, groups=1, bias=True, dilation=1, padding_mode='zeros', step_mode='s')[源代码]#

基类：ConvTranspose2d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.ConvTranspose2d

  • English

参数:

Refer to torch.nn.ConvTranspose2d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.ConvTranspose3d(in_channels, out_channels, kernel_size, stride=1, padding=0, output_padding=0, groups=1, bias=True, dilation=1, padding_mode='zeros', step_mode='s')[源代码]#

基类：ConvTranspose3d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.ConvTranspose3d

  • English

参数:

Refer to torch.nn.ConvTranspose3d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.GroupNorm(num_groups, num_channels, eps=1e-05, affine=True, step_mode='s')[源代码]#

基类：GroupNorm, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • num_groups (int)

  • num_channels (int)

  • eps (float)

  • affine (bool)

其他的参数API参见 torch.nn.GroupNorm

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • num_groups (int)

  • num_channels (int)

  • eps (float)

  • affine (bool)

Refer to torch.nn.GroupNorm for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.MaxPool1d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False, step_mode='s')[源代码]#

基类：MaxPool1d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.MaxPool1d

  • English

参数:

Refer to torch.nn.MaxPool1d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.MaxPool2d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False, step_mode='s')[源代码]#

基类：MaxPool2d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.MaxPool2d

  • English

参数:

Refer to torch.nn.MaxPool2d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.MaxPool3d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False, step_mode='s')[源代码]#

基类：MaxPool3d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.MaxPool3d

  • English

参数:

Refer to torch.nn.MaxPool3d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.AvgPool1d(kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=True, step_mode='s')[源代码]#

基类：AvgPool1d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.AvgPool1d

  • English

参数:

Refer to torch.nn.AvgPool1d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.AvgPool2d(kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=True, divisor_override=None, step_mode='s')[源代码]#

基类：AvgPool2d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.AvgPool2d

  • English

参数:

Refer to torch.nn.AvgPool2d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.AvgPool3d(kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=True, divisor_override=None, step_mode='s')[源代码]#

基类：AvgPool3d, StepModule

API Language - 中文 | English

  • 中文

参数:

其他的参数API参见 torch.nn.AvgPool3d

  • English

参数:

Refer to torch.nn.AvgPool3d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.AdaptiveAvgPool1d(output_size, step_mode='s')[源代码]#

基类：AdaptiveAvgPool1d, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • output_size (int | tuple[int])

其他的参数API参见 torch.nn.AdaptiveAvgPool1d

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • output_size (int | tuple[int])

Refer to torch.nn.AdaptiveAvgPool1d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.AdaptiveAvgPool2d(output_size, step_mode='s')[源代码]#

基类：AdaptiveAvgPool2d, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • output_size (int | None | tuple[int | None, int | None])

其他的参数API参见 torch.nn.AdaptiveAvgPool2d

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • output_size (int | None | tuple[int | None, int | None])

Refer to torch.nn.AdaptiveAvgPool2d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.AdaptiveAvgPool3d(output_size, step_mode='s')[源代码]#

基类：AdaptiveAvgPool3d, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • output_size (int | None | tuple[int | None, int | None, int | None])

其他的参数API参见 torch.nn.AdaptiveAvgPool3d

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • output_size (int | None | tuple[int | None, int | None, int | None])

Refer to torch.nn.AdaptiveAvgPool3d for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.Linear(in_features, out_features, bias=True, step_mode='s')[源代码]#

基类：Linear, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • in_features (int)

  • out_features (int)

  • bias (bool)

其他的参数API参见 torch.nn.Linear

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • in_features (int)

  • out_features (int)

  • bias (bool)

Refer to torch.nn.Linear for other parameters' API

class spikingjelly.activation_based.layer.stateless_wrapper.Flatten(start_dim=1, end_dim=-1, step_mode='s')[源代码]#

基类：Flatten, StepModule

API Language - 中文 | English

  • 中文

参数:

  • step_mode (str) -- 步进模式，可以为 's' (单步) 或 'm' (多步)

  • start_dim (int)

  • end_dim (int)

其他的参数API参见 torch.nn.Flatten

  • English

参数:

  • step_mode (str) -- the step mode, which can be s (single-step) or m (multi-step)

  • start_dim (int)

  • end_dim (int)

Refer to torch.nn.Flatten for other parameters' API

forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.WSConv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', step_mode='s')[源代码]#

基类：Conv2d

API Language - 中文 | English

  • 中文

参数:

  • gain (bool) -- 是否对权重引入可学习的缩放系数

  • eps (float) -- 预防数值问题的小量

  • in_channels (int)

  • out_channels (int)

  • kernel_size (tuple[int, ...])

  • stride (tuple[int, ...])

  • padding (str | tuple[int, ...])

  • dilation (tuple[int, ...])

  • groups (int)

  • bias (Tensor | None)

  • padding_mode (str)

  • step_mode (str)

其他的参数API参见 Conv2d

  • English

参数:

  • gain -- whether introduce learnable scale factors for weights

  • eps (float) -- a small number to prevent numerical problems

  • in_channels (int)

  • out_channels (int)

  • kernel_size (tuple[int, ...])

  • stride (tuple[int, ...])

  • padding (str | tuple[int, ...])

  • dilation (tuple[int, ...])

  • groups (int)

  • bias (Tensor | None)

  • padding_mode (str)

  • step_mode (bool)

Refer to Conv2d for other parameters' API

get_weight()[源代码]#
forward(x)[源代码]#
参数:

x (Tensor)

class spikingjelly.activation_based.layer.stateless_wrapper.WSLinear(in_features, out_features, bias=True, step_mode='s', gain=True, eps=0.0001)[源代码]#

基类：Linear

API Language - 中文 | English

  • 中文

参数:

  • gain (bool) -- 是否对权重引入可学习的缩放系数

  • eps (float) -- 预防数值问题的小量

  • in_features (int)

  • out_features (int)

  • bias (bool)

其他的参数API参见 Linear

  • English

参数:

  • gain -- whether introduce learnable scale factors for weights

  • eps (float) -- a small number to prevent numerical problems

  • in_features (int)

  • out_features (int)

  • bias (bool)

Refer to Linear for other parameters' API

get_weight()[源代码]#
forward(x)[源代码]#
参数:

x (Tensor)