Layers

Helper functions used to build PyTorch timeseries models.

---

source

test_module_to_torchscript

 test_module_to_torchscript (m:torch.nn.modules.module.Module,
                             inputs:torch.Tensor, trace:bool=True,
                             script:bool=True, serialize:bool=True,
                             verbose:bool=True)

Tests if a PyTorch module can be correctly traced or scripted and serialized

	Type	Default	Details
m	Module		The PyTorch module to be tested.
inputs	Tensor		A tensor or tuple of tensors representing the inputs to the model.
trace	bool	True	If True, attempts to trace the model. Defaults to True.
script	bool	True	If True, attempts to script the model. Defaults to True.
serialize	bool	True	If True, saves and loads the traced/scripted module to ensure it can be serialized. Defaults to True.
verbose	bool	True	If True, prints detailed information about the tracing and scripting process. Defaults to True.

m = nn.Linear(10, 2)
inp = torch.randn(3, 10)
test_module_to_torchscript(m, inp, trace=True, script=True, serialize=True, verbose=True)

output.shape: torch.Size([3, 2])
Tracing...
...Linear has been successfully traced 😃

True

---

source

init_lin_zero

 init_lin_zero (m)

---

source

SwishBeta

 SwishBeta ()

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

SmeLU

 SmeLU (beta:float=2.0)

Smooth ReLU activation function based on https://arxiv.org/pdf/2202.06499.pdf

	Type	Default	Details
beta	float	2.0	Beta value
Returns	None

---

source

Chomp1d

 Chomp1d (chomp_size)

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing to nest them in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will have their parameters converted too when you call :meth:to, etc.

.. note:: As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

---

source

SameConv1d

 SameConv1d (ni, nf, ks=3, stride=1, dilation=1, **kwargs)

Conv1d with padding=‘same’

---

source

Pad1d

 Pad1d (padding, value=0.0)

Pads the input tensor boundaries with a constant value.

For N-dimensional padding, use :func:torch.nn.functional.pad().

Args: padding (int, tuple): the size of the padding. If is int, uses the same padding in both boundaries. If a 2-tuple, uses (:math:\text{padding\_left}, :math:\text{padding\_right})

Shape: - Input: :math:(C, W_{in}) or :math:(N, C, W_{in}). - Output: :math:(C, W_{out}) or :math:(N, C, W_{out}), where

  :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`

Examples::

>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> m = nn.ConstantPad1d(2, 3.5)
>>> input = torch.randn(1, 2, 4)
>>> input
tensor([[[-1.0491, -0.7152, -0.0749,  0.8530],
         [-1.3287,  1.8966,  0.1466, -0.2771]]])
>>> m(input)
tensor([[[ 3.5000,  3.5000, -1.0491, -0.7152, -0.0749,  0.8530,  3.5000,
           3.5000],
         [ 3.5000,  3.5000, -1.3287,  1.8966,  0.1466, -0.2771,  3.5000,
           3.5000]]])
>>> m = nn.ConstantPad1d(2, 3.5)
>>> input = torch.randn(1, 2, 3)
>>> input
tensor([[[ 1.6616,  1.4523, -1.1255],
         [-3.6372,  0.1182, -1.8652]]])
>>> m(input)
tensor([[[ 3.5000,  3.5000,  1.6616,  1.4523, -1.1255,  3.5000,  3.5000],
         [ 3.5000,  3.5000, -3.6372,  0.1182, -1.8652,  3.5000,  3.5000]]])
>>> # using different paddings for different sides
>>> m = nn.ConstantPad1d((3, 1), 3.5)
>>> m(input)
tensor([[[ 3.5000,  3.5000,  3.5000,  1.6616,  1.4523, -1.1255,  3.5000],
         [ 3.5000,  3.5000,  3.5000, -3.6372,  0.1182, -1.8652,  3.5000]]])

---

source

same_padding1d

 same_padding1d (seq_len, ks, stride=1, dilation=1)

Same padding formula as used in Tensorflow

---

source

Conv2d

 Conv2d (ni, nf, kernel_size=None, ks=None, stride=1, padding='same',
         dilation=1, init='auto', bias_std=0.01, **kwargs)

conv1d layer with padding=‘same’, ‘valid’, or any integer (defaults to ‘same’)

---

source

Conv2dSame

 Conv2dSame (ni, nf, ks=(3, 3), stride=(1, 1), dilation=(1, 1), **kwargs)

Conv2d with padding=‘same’

---

source

Pad2d

 Pad2d (padding, value=0.0)

Pads the input tensor boundaries with a constant value.

For N-dimensional padding, use :func:torch.nn.functional.pad().

Args: padding (int, tuple): the size of the padding. If is int, uses the same padding in all boundaries. If a 4-tuple, uses (:math:\text{padding\_left}, :math:\text{padding\_right}, :math:\text{padding\_top}, :math:\text{padding\_bottom})

Shape: - Input: :math:(N, C, H_{in}, W_{in}) or :math:(C, H_{in}, W_{in}). - Output: :math:(N, C, H_{out}, W_{out}) or :math:(C, H_{out}, W_{out}), where

  :math:`H_{out} = H_{in} + \text{padding\_top} + \text{padding\_bottom}`

  :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`

Examples::

>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> m = nn.ConstantPad2d(2, 3.5)
>>> input = torch.randn(1, 2, 2)
>>> input
tensor([[[ 1.6585,  0.4320],
         [-0.8701, -0.4649]]])
>>> m(input)
tensor([[[ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000,  3.5000],
         [ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000,  3.5000],
         [ 3.5000,  3.5000,  1.6585,  0.4320,  3.5000,  3.5000],
         [ 3.5000,  3.5000, -0.8701, -0.4649,  3.5000,  3.5000],
         [ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000,  3.5000],
         [ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000,  3.5000]]])
>>> # using different paddings for different sides
>>> m = nn.ConstantPad2d((3, 0, 2, 1), 3.5)
>>> m(input)
tensor([[[ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000],
         [ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000],
         [ 3.5000,  3.5000,  3.5000,  1.6585,  0.4320],
         [ 3.5000,  3.5000,  3.5000, -0.8701, -0.4649],
         [ 3.5000,  3.5000,  3.5000,  3.5000,  3.5000]]])

---

source

same_padding2d

 same_padding2d (H, W, ks, stride=(1, 1), dilation=(1, 1))

Same padding formula as used in Tensorflow

bs = 2
c_in = 3
c_out = 5
h = 16
w = 20
t = torch.rand(bs, c_in, h, w)
test_eq(Conv2dSame(c_in, c_out, ks=3, stride=1, dilation=1, bias=False)(t).shape, (bs, c_out, h, w))
test_eq(Conv2dSame(c_in, c_out, ks=(3, 1), stride=1, dilation=1, bias=False)(t).shape, (bs, c_out, h, w))
test_eq(Conv2dSame(c_in, c_out, ks=3, stride=(1, 1), dilation=(2, 2), bias=False)(t).shape, (bs, c_out, h, w))
test_eq(Conv2dSame(c_in, c_out, ks=3, stride=(2, 2), dilation=(1, 1), bias=False)(t).shape, (bs, c_out, h//2, w//2))
test_eq(Conv2dSame(c_in, c_out, ks=3, stride=(2, 2), dilation=(2, 2), bias=False)(t).shape, (bs, c_out, h//2, w//2))
test_eq(Conv2d(c_in, c_out, ks=3, padding='same', stride=1, dilation=1, bias=False)(t).shape, (bs, c_out, h, w))

---

source

CausalConv1d

 CausalConv1d (ni, nf, ks, stride=1, dilation=1, groups=1, bias=True)

Applies a 1D convolution over an input signal composed of several input planes.

In the simplest case, the output value of the layer with input size :math:(N, C_{\text{in}}, L) and output :math:(N, C_{\text{out}}, L_{\text{out}}) can be precisely described as:

.. math:: (N_i, C_{j}) = (C{j}) + {k = 0}^{C_{in} - 1} (C_{_j}, k) (N_i, k)

where :math:\star is the valid cross-correlation_ operator, :math:N is a batch size, :math:C denotes a number of channels, :math:L is a length of signal sequence.

This module supports :ref:TensorFloat32<tf32_on_ampere>.

On certain ROCm devices, when using float16 inputs this module will use :ref:different precision<fp16_on_mi200> for backward.

• :attr:stride controls the stride for the cross-correlation, a single number or a one-element tuple.

• :attr:padding controls the amount of padding applied to the input. It can be either a string {‘valid’, ‘same’} or a tuple of ints giving the amount of implicit padding applied on both sides.

• :attr:dilation controls the spacing between the kernel points; also known as the à trous algorithm. It is harder to describe, but this link_ has a nice visualization of what :attr:dilation does.

• :attr:groups controls the connections between inputs and outputs. :attr:in_channels and :attr:out_channels must both be divisible by :attr:groups. For example,

  • At groups=1, all inputs are convolved to all outputs.
  • At groups=2, the operation becomes equivalent to having two conv layers side by side, each seeing half the input channels and producing half the output channels, and both subsequently concatenated.
  • At groups= :attr:in_channels, each input channel is convolved with its own set of filters (of size :math:\frac{\text{out\_channels}}{\text{in\_channels}}).

Note: When groups == in_channels and out_channels == K * in_channels, where K is a positive integer, this operation is also known as a “depthwise convolution”.

In other words, for an input of size :math:`(N, C_{in}, L_{in})`,
a depthwise convolution with a depthwise multiplier `K` can be performed with the arguments
:math:`(C_\text{in}=C_\text{in}, C_\text{out}=C_\text{in} \times \text{K}, ..., \text{groups}=C_\text{in})`.

Note: In some circumstances when given tensors on a CUDA device and using CuDNN, this operator may select a nondeterministic algorithm to increase performance. If this is undesirable, you can try to make the operation deterministic (potentially at a performance cost) by setting torch.backends.cudnn.deterministic = True. See :doc:/notes/randomness for more information.

Note: padding='valid' is the same as no padding. padding='same' pads the input so the output has the shape as the input. However, this mode doesn’t support any stride values other than 1.

Note: This module supports complex data types i.e. complex32, complex64, complex128.

Args: in_channels (int): Number of channels in the input image out_channels (int): Number of channels produced by the convolution kernel_size (int or tuple): Size of the convolving kernel stride (int or tuple, optional): Stride of the convolution. Default: 1 padding (int, tuple or str, optional): Padding added to both sides of the input. Default: 0 padding_mode (str, optional): 'zeros', 'reflect', 'replicate' or 'circular'. Default: 'zeros' dilation (int or tuple, optional): Spacing between kernel elements. Default: 1 groups (int, optional): Number of blocked connections from input channels to output channels. Default: 1 bias (bool, optional): If True, adds a learnable bias to the output. Default: True

Shape: - Input: :math:(N, C_{in}, L_{in}) or :math:(C_{in}, L_{in}) - Output: :math:(N, C_{out}, L_{out}) or :math:(C_{out}, L_{out}), where

  .. math::
      L_{out} = \left\lfloor\frac{L_{in} + 2 \times \text{padding} - \text{dilation}
                \times (\text{kernel\_size} - 1) - 1}{\text{stride}} + 1\right\rfloor

Attributes: weight (Tensor): the learnable weights of the module of shape :math:(\text{out\_channels}, \frac{\text{in\_channels}}{\text{groups}}, \text{kernel\_size}). The values of these weights are sampled from :math:\mathcal{U}(-\sqrt{k}, \sqrt{k}) where :math:k = \frac{groups}{C_\text{in} * \text{kernel\_size}} bias (Tensor): the learnable bias of the module of shape (out_channels). If :attr:bias is True, then the values of these weights are sampled from :math:\mathcal{U}(-\sqrt{k}, \sqrt{k}) where :math:k = \frac{groups}{C_\text{in} * \text{kernel\_size}}

Examples::

>>> m = nn.Conv1d(16, 33, 3, stride=2)
>>> input = torch.randn(20, 16, 50)
>>> output = m(input)

.. _cross-correlation: https://en.wikipedia.org/wiki/Cross-correlation

.. _link: https://github.com/vdumoulin/conv_arithmetic/blob/master/README.md

---

source

Conv1d

 Conv1d (ni, nf, kernel_size=None, ks=None, stride=1, padding='same',
         dilation=1, init='auto', bias_std=0.01, **kwargs)

conv1d layer with padding=‘same’, ‘causal’, ‘valid’, or any integer (defaults to ‘same’)

bs = 2
c_in = 3
c_out = 5
seq_len = 512
t = torch.rand(bs, c_in, seq_len)
dilation = 1
test_eq(CausalConv1d(c_in, c_out, ks=3, dilation=dilation)(t).shape, Conv1d(c_in, c_out, ks=3, padding="same", dilation=dilation)(t).shape)
dilation = 2
test_eq(CausalConv1d(c_in, c_out, ks=3, dilation=dilation)(t).shape, Conv1d(c_in, c_out, ks=3, padding="same", dilation=dilation)(t).shape)

bs = 2
ni = 3
nf = 5
seq_len = 6
ks = 3
t = torch.rand(bs, c_in, seq_len)
test_eq(Conv1d(ni, nf, ks, padding=0)(t).shape, (bs, c_out, seq_len - (2 * (ks//2))))
test_eq(Conv1d(ni, nf, ks, padding='valid')(t).shape, (bs, c_out, seq_len - (2 * (ks//2))))
test_eq(Conv1d(ni, nf, ks, padding='same')(t).shape, (bs, c_out, seq_len))
test_eq(Conv1d(ni, nf, ks, padding='causal')(t).shape, (bs, c_out, seq_len))
test_error('use kernel_size or ks but not both simultaneously', Conv1d, ni, nf, kernel_size=3, ks=3)
test_error('you need to pass a ks', Conv1d, ni, nf)

conv = Conv1d(ni, nf, ks, padding='same')
init_linear(conv, None, init='auto', bias_std=.01)
conv

Conv1d(3, 5, kernel_size=(3,), stride=(1,), padding=(1,))

conv = Conv1d(ni, nf, ks, padding='causal')
init_linear(conv, None, init='auto', bias_std=.01)
conv

CausalConv1d(3, 5, kernel_size=(3,), stride=(1,))

conv = Conv1d(ni, nf, ks, padding='valid')
init_linear(conv, None, init='auto', bias_std=.01)
weight_norm(conv)
conv

Conv1d(3, 5, kernel_size=(3,), stride=(1,))

conv = Conv1d(ni, nf, ks, padding=0)
init_linear(conv, None, init='auto', bias_std=.01)
weight_norm(conv)
conv

Conv1d(3, 5, kernel_size=(3,), stride=(1,))

---

source

SeparableConv1d

 SeparableConv1d (ni, nf, ks, stride=1, padding='same', dilation=1,
                  bias=True, bias_std=0.01)

Same as nn.Module, but no need for subclasses to call super().__init__

bs = 64
c_in = 6
c_out = 5
seq_len = 512
t = torch.rand(bs, c_in, seq_len)
test_eq(SeparableConv1d(c_in, c_out, 3)(t).shape, (bs, c_out, seq_len))

---

source

AddCoords1d

 AddCoords1d ()

Add coordinates to ease position identification without modifying mean and std

bs = 2
c_in = 3
c_out = 5
seq_len = 50

t = torch.rand(bs, c_in, seq_len)
t = (t - t.mean()) / t.std()
test_eq(AddCoords1d()(t).shape, (bs, c_in + 1, seq_len))
new_t = AddCoords1d()(t)
test_close(new_t.mean(),0, 1e-2)
test_close(new_t.std(), 1, 1e-2)

---

source

ConvBlock

 ConvBlock (ni, nf, kernel_size=None, ks=3, stride=1, padding='same',
            bias=None, bias_std=0.01, norm='Batch', zero_norm=False,
            bn_1st=True, act=<class 'torch.nn.modules.activation.ReLU'>,
            act_kwargs={}, init='auto', dropout=0.0, xtra=None,
            coord=False, separable=False, **kwargs)

Create a sequence of conv1d (ni to nf), activation (if act_cls) and norm_type layers.

---

source

ResBlock1dPlus

 ResBlock1dPlus (expansion, ni, nf, coord=False, stride=1, groups=1,
                 reduction=None, nh1=None, nh2=None, dw=False, g2=1,
                 sa=False, sym=False, norm='Batch', zero_norm=True,
                 act_cls=<class 'torch.nn.modules.activation.ReLU'>, ks=3,
                 pool=<function AvgPool>, pool_first=True, **kwargs)

Resnet block from ni to nh with stride

---

source

SEModule1d

 SEModule1d (ni, reduction=16, act=<class
             'torch.nn.modules.activation.ReLU'>, act_kwargs={})

Squeeze and excitation module for 1d

t = torch.rand(8, 32, 12)
test_eq(SEModule1d(t.shape[1], 16, act=nn.ReLU, act_kwargs={})(t).shape, t.shape)

---

source

Norm

 Norm (nf, ndim=1, norm='Batch', zero_norm=False, init=True, **kwargs)

Norm layer with nf features and ndim with auto init.

bs = 2
ni = 3
nf = 5
sl = 4
ks = 5

t = torch.rand(bs, ni, sl)
test_eq(ConvBlock(ni, nf, ks)(t).shape, (bs, nf, sl))
test_eq(ConvBlock(ni, nf, ks, padding='causal')(t).shape, (bs, nf, sl))
test_eq(ConvBlock(ni, nf, ks, coord=True)(t).shape, (bs, nf, sl))

test_eq(BN1d(ni)(t).shape, (bs, ni, sl))
test_eq(BN1d(ni).weight.data.mean().item(), 1.)
test_eq(BN1d(ni, zero_norm=True).weight.data.mean().item(), 0.)

test_eq(ConvBlock(ni, nf, ks, norm='batch', zero_norm=True)[1].weight.data.unique().item(), 0)
test_ne(ConvBlock(ni, nf, ks, norm='batch', zero_norm=False)[1].weight.data.unique().item(), 0)
test_eq(ConvBlock(ni, nf, ks, bias=False)[0].bias, None)
ConvBlock(ni, nf, ks, act=Swish, coord=True)

ConvBlock(
  (0): AddCoords1d()
  (1): Conv1d(4, 5, kernel_size=(5,), stride=(1,), padding=(2,), bias=False)
  (2): BatchNorm1d(5, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (3): Swish()
)

---

source

LinLnDrop

 LinLnDrop (n_in, n_out, ln=True, p=0.0, act=None, lin_first=False)

Module grouping LayerNorm1d, Dropout and Linear layers

LinLnDrop(2, 3, p=.5)

LinLnDrop(
  (0): LayerNorm((2,), eps=1e-05, elementwise_affine=True)
  (1): Dropout(p=0.5, inplace=False)
  (2): Linear(in_features=2, out_features=3, bias=False)
)

---

source

LambdaPlus

 LambdaPlus (func, *args, **kwargs)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

ReZero

 ReZero (module)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Clip

 Clip (min=None, max=None)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Clamp

 Clamp (min=None, max=None)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

SoftMax

 SoftMax (dim=-1)

SoftMax layer

---

source

LastStep

 LastStep ()

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Max

 Max (dim=None, keepdim=False)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Reshape

 Reshape (*shape)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

View

 View (*shape)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Transpose

 Transpose (*dims, contiguous=False)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Permute

 Permute (*dims)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Unfold

 Unfold (dim, size, step=1)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Concat

 Concat (dim=1)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Add

 Add ()

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Unsqueeze

 Unsqueeze (dim=-1)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

Squeeze

 Squeeze (dim=-1)

Same as nn.Module, but no need for subclasses to call super().__init__

bs = 2
nf = 5
sl = 4

t = torch.rand(bs, nf, sl)
test_eq(Permute(0,2,1)(t).shape, (bs, sl, nf))
test_eq(Max(1)(t).shape, (bs, sl))
test_eq(Transpose(1,2)(t).shape, (bs, sl, nf))
test_eq(Transpose(1,2, contiguous=True)(t).shape, (bs, sl, nf))
test_eq(View(-1, 2, 10)(t).shape, (bs, 1, 2, 10))
test_eq(Reshape(-1, 2, 10)(t).shape, (bs, 1, 2, 10))
test_eq(Reshape()(t).shape, (2, 20))
test_eq(Reshape(-1)(t).shape, (40,))
Transpose(1,2), Permute(0,2,1), View(-1, 2, 10), Transpose(1,2, contiguous=True), Reshape(-1, 2, 10), Noop

(Transpose(1, 2),
 Permute(dims=0, 2, 1),
 View(bs, -1, 2, 10),
 Transpose(dims=1, 2).contiguous(),
 Reshape(bs, -1, 2, 10),
 Sequential())

---

source

DropPath

 DropPath (p=None)

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

It’s similar to Dropout but it drops individual connections instead of nodes. Original code in https://github.com/rwightman/pytorch-image-models (timm library)

t = torch.ones(100,2,3)
test_eq(DropPath(0.)(t), t)
assert DropPath(0.5)(t).max() >= 1

---

source

Sharpen

 Sharpen (T=0.5)

This is used to increase confidence in predictions - MixMatch paper

n_samples = 1000
n_classes = 3

t = (torch.rand(n_samples, n_classes) - .5) * 10
probas = F.softmax(t, -1)
sharpened_probas = Sharpen()(probas)
plt.plot(probas.flatten().sort().values, color='r')
plt.plot(sharpened_probas.flatten().sort().values, color='b')
plt.show()
test_gt(sharpened_probas[n_samples//2:].max(-1).values.sum().item(), probas[n_samples//2:].max(-1).values.sum().item())

---

source

Sequential

 Sequential (*args)

Class that allows you to pass one or multiple inputs

---

source

TimeDistributed

 TimeDistributed (module, batch_first=False)

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing to nest them in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will have their parameters converted too when you call :meth:to, etc.

.. note:: As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

---

source

get_calibrator

 get_calibrator (calibrator=None, n_classes=1, **kwargs)

---

source

Matrix_Scale

 Matrix_Scale (n_classes=1, dirichlet=False)

Used to perform Matrix Scaling (dirichlet=False) or Dirichlet calibration (dirichlet=True)

---

source

Vector_Scale

 Vector_Scale (n_classes=1, dirichlet=False)

Used to perform Vector Scaling (dirichlet=False) or Diagonal Dirichlet calibration (dirichlet=True)

---

source

Temp_Scale

 Temp_Scale (temp=1.0, dirichlet=False)

Used to perform Temperature Scaling (dirichlet=False) or Single-parameter Dirichlet calibration (dirichlet=True)

bs = 2
c_out = 3

t = torch.rand(bs, c_out)
for calibrator, cal_name in zip(['temp', 'vector', 'matrix'], ['Temp_Scale', 'Vector_Scale', 'Matrix_Scale']): 
    cal = get_calibrator(calibrator, n_classes=c_out)
#     print(calibrator)
#     print(cal.weight, cal.bias, '\n')
    test_eq(cal(t), t)
    test_eq(cal.__class__.__name__, cal_name)
for calibrator, cal_name in zip(['dtemp', 'dvector', 'dmatrix'], ['Temp_Scale', 'Vector_Scale', 'Matrix_Scale']):
    cal = get_calibrator(calibrator, n_classes=c_out)
#     print(calibrator)
#     print(cal.weight, cal.bias, '\n')
    test_eq(cal(t), F.log_softmax(t, dim=1))
    test_eq(cal.__class__.__name__, cal_name)

bs = 2
c_out = 3

t = torch.rand(bs, c_out)

test_eq(Temp_Scale()(t).shape, t.shape)
test_eq(Vector_Scale(c_out)(t).shape, t.shape)
test_eq(Matrix_Scale(c_out)(t).shape, t.shape)
test_eq(Temp_Scale(dirichlet=True)(t).shape, t.shape)
test_eq(Vector_Scale(c_out, dirichlet=True)(t).shape, t.shape)
test_eq(Matrix_Scale(c_out, dirichlet=True)(t).shape, t.shape)

test_eq(Temp_Scale()(t), t)
test_eq(Vector_Scale(c_out)(t), t)
test_eq(Matrix_Scale(c_out)(t), t)

bs = 2
c_out = 5

t = torch.rand(bs, c_out)
test_eq(Vector_Scale(c_out)(t), t)
test_eq(Vector_Scale(c_out).weight.data, torch.ones(c_out))
test_eq(Vector_Scale(c_out).weight.requires_grad, True)
test_eq(type(Vector_Scale(c_out).weight), torch.nn.parameter.Parameter)

bs = 2
c_out = 3
weight = 2
bias = 1

t = torch.rand(bs, c_out)
test_eq(Matrix_Scale(c_out)(t).shape, t.shape)
test_eq(Matrix_Scale(c_out).weight.requires_grad, True)
test_eq(type(Matrix_Scale(c_out).weight), torch.nn.parameter.Parameter)

---

source

LogitAdjustmentLayer

 LogitAdjustmentLayer (class_priors)

Logit Adjustment for imbalanced datasets

bs, n_classes = 16, 3
class_priors = torch.rand(n_classes)
logits = torch.randn(bs, n_classes) * 2
test_eq(LogitAdjLayer(class_priors)(logits), logits + class_priors)

---

source

MaxPPVPool1d

 MaxPPVPool1d ()

Drop-in replacement for AdaptiveConcatPool1d - multiplies nf by 2

---

source

PPAuc

 PPAuc (dim=-1)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

PPV

 PPV (dim=-1)

Same as nn.Module, but no need for subclasses to call super().__init__

bs = 2
nf = 5
sl = 4

t = torch.rand(bs, nf, sl)
test_eq(MaxPPVPool1d()(t).shape, (bs, nf*2, 1))
test_eq(MaxPPVPool1d()(t).shape, AdaptiveConcatPool1d(1)(t).shape)

---

source

AdaptiveWeightedAvgPool1d

 AdaptiveWeightedAvgPool1d (n_in, seq_len, mult=2, n_layers=2, ln=False,
                            dropout=0.5, act=ReLU(), zero_init=True)

Global Pooling layer that performs a weighted average along the temporal axis

It can be considered as a channel-wise form of local temporal attention. Inspired by the paper: Hyun, J., Seong, H., & Kim, E. (2019). Universal Pooling–A New Pooling Method for Convolutional Neural Networks. arXiv preprint arXiv:1907.11440.

---

source

GAWP1d

 GAWP1d (n_in, seq_len, n_layers=2, ln=False, dropout=0.5, act=ReLU(),
         zero_init=False)

Global AdaptiveWeightedAvgPool1d + Flatten

---

source

GACP1d

 GACP1d (output_size=1)

Global AdaptiveConcatPool + Flatten

---

source

GAP1d

 GAP1d (output_size=1)

Global Adaptive Pooling + Flatten

---

source

gwa_pool_head

 gwa_pool_head (n_in, c_out, seq_len, bn=True, fc_dropout=0.0)

---

source

GlobalWeightedAveragePool1d

 GlobalWeightedAveragePool1d (n_in, seq_len)

Global Weighted Average Pooling layer

Inspired by Building Efficient CNN Architecture for Offline Handwritten Chinese Character Recognition https://arxiv.org/pdf/1804.01259.pdf

t = torch.randn(16, 64, 50)
head = gwa_pool_head(64, 5, 50)
test_eq(head(t).shape, (16, 5))

---

source

attentional_pool_head

 attentional_pool_head (n_in, c_out, seq_len=None, bn=True, **kwargs)

---

source

GAttP1d

 GAttP1d (n_in, c_out, bn=False)

A sequential container.

Modules will be added to it in the order they are passed in the constructor. Alternatively, an OrderedDict of modules can be passed in. The forward() method of [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) accepts any input and forwards it to the first module it contains. It then “chains” outputs to inputs sequentially for each subsequent module, finally returning the output of the last module.

The value a [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) provides over manually calling a sequence of modules is that it allows treating the whole container as a single module, such that performing a transformation on the [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) applies to each of the modules it stores (which are each a registered submodule of the [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential)).

What’s the difference between a [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) and a :class:torch.nn.ModuleList? A ModuleList is exactly what it sounds like–a list for storing Module s! On the other hand, the layers in a [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) are connected in a cascading way.

Example::

# Using Sequential to create a small model. When `model` is run,
# input will first be passed to `Conv2d(1,20,5)`. The output of
# `Conv2d(1,20,5)` will be used as the input to the first
# `ReLU`; the output of the first `ReLU` will become the input
# for `Conv2d(20,64,5)`. Finally, the output of
# `Conv2d(20,64,5)` will be used as input to the second `ReLU`
model = nn.Sequential(
          nn.Conv2d(1,20,5),
          nn.ReLU(),
          nn.Conv2d(20,64,5),
          nn.ReLU()
        )

# Using Sequential with OrderedDict. This is functionally the
# same as the above code
model = nn.Sequential(OrderedDict([
          ('conv1', nn.Conv2d(1,20,5)),
          ('relu1', nn.ReLU()),
          ('conv2', nn.Conv2d(20,64,5)),
          ('relu2', nn.ReLU())
        ]))

---

source

AttentionalPool1d

 AttentionalPool1d (n_in, c_out, bn=False)

Global Adaptive Pooling layer inspired by Attentional Pooling for Action Recognition https://arxiv.org/abs/1711.01467

bs, c_in, seq_len = 16, 1, 50
c_out = 3
t = torch.rand(bs, c_in, seq_len)
test_eq(GAP1d()(t).shape, (bs, c_in))
test_eq(GACP1d()(t).shape, (bs, c_in*2))
bs, c_in, seq_len = 16, 4, 50
t = torch.rand(bs, c_in, seq_len)
test_eq(GAP1d()(t).shape, (bs, c_in))
test_eq(GACP1d()(t).shape, (bs, c_in*2))
test_eq(GAWP1d(c_in, seq_len, n_layers=2, ln=False, dropout=0.5, act=nn.ReLU(), zero_init=False)(t).shape, (bs, c_in))
test_eq(GAWP1d(c_in, seq_len, n_layers=2, ln=False, dropout=0.5, act=nn.ReLU(), zero_init=False)(t).shape, (bs, c_in))
test_eq(GAWP1d(c_in, seq_len, n_layers=1, ln=False, dropout=0.5, zero_init=False)(t).shape, (bs, c_in))
test_eq(GAWP1d(c_in, seq_len, n_layers=1, ln=False, dropout=0.5, zero_init=True)(t).shape, (bs, c_in))
test_eq(AttentionalPool1d(c_in, c_out)(t).shape, (bs, c_out, 1))

bs, c_in, seq_len = 16, 128, 50
c_out = 14
t = torch.rand(bs, c_in, seq_len)
attp = attentional_pool_head(c_in, c_out)
test_eq(attp(t).shape, (bs, c_out))

---

source

PoolingLayer

 PoolingLayer (method='cls', seq_len=None, token=True, seq_last=True)

Same as nn.Module, but no need for subclasses to call super().__init__

t = torch.arange(24).reshape(2, 3, 4).float()
test_eq(PoolingLayer('cls', token=True, seq_last=True)(t), tensor([[ 0.,  4.,  8.], [12., 16., 20.]]))
test_eq(PoolingLayer('max', token=True, seq_last=True)(t), tensor([[ 3.,  7., 11.], [15., 19., 23.]]))
test_close(PoolingLayer('mean', token=True, seq_last=True)(t), tensor([[ 2.,  6., 10.], [14., 18., 22.]]))
test_close(PoolingLayer('max-mean', token=True, seq_last=True)(t), tensor([[ 3.,  7., 11.,  2.,  6., 10.],
                                                                           [15., 19., 23., 14., 18., 22.]]))
test_close(PoolingLayer('flatten', token=True, seq_last=True)(t), tensor([[ 1.,  2.,  3.,  5.,  6.,  7.,  9., 10., 11.],
                                                                          [13., 14., 15., 17., 18., 19., 21., 22., 23.]]))
test_eq(PoolingLayer('linear', seq_len=4, token=True, seq_last=True)(t).shape, (2, 3))
test_eq(PoolingLayer('max', token=False, seq_last=True)(t), tensor([[ 3.,  7., 11.], [15., 19., 23.]]))
test_close(PoolingLayer('mean', token=False, seq_last=True)(t), tensor([[ 1.5000,  5.5000,  9.5000],
                                                                        [13.5000, 17.5000, 21.5000]]))
test_close(PoolingLayer('max-mean', token=False, seq_last=True)(t), tensor([[ 3.,  7., 11.,  1.5000,  5.5000,  9.5000],
                                                                            [15., 19., 23., 13.5000, 17.5000, 21.5000]]))
test_close(PoolingLayer('flatten', token=False, seq_last=True)(t), tensor([[ 0.,  1.,  2.,  3.,  4.,  5.,  6.,  7.,  8.,  9., 10., 11.],
                                                                           [12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23.]]))
test_eq(PoolingLayer('linear', seq_len=4, token=False, seq_last=True)(t).shape, (2, 3))

t = torch.arange(24).reshape(2, 3, 4).swapaxes(1,2).float()
test_eq(PoolingLayer('cls', token=True, seq_last=False)(t), tensor([[ 0.,  4.,  8.], [12., 16., 20.]]))
test_eq(PoolingLayer('max', token=True, seq_last=False)(t), tensor([[ 3.,  7., 11.], [15., 19., 23.]]))
test_close(PoolingLayer('mean', token=True, seq_last=False)(t), tensor([[ 2.,  6., 10.], [14., 18., 22.]]))
test_close(PoolingLayer('max-mean', token=True, seq_last=False)(t), tensor([[ 3.,  7., 11.,  2.,  6., 10.],
                                                                           [15., 19., 23., 14., 18., 22.]]))
test_close(PoolingLayer('flatten', token=True, seq_last=False)(t), tensor([[ 1.,  5.,  9.,  2.,  6., 10.,  3.,  7., 11.],
                                                                           [13., 17., 21., 14., 18., 22., 15., 19., 23.]]))
t = torch.arange(24).reshape(2, 3, 4).swapaxes(1,2).float()
test_eq(PoolingLayer('conv1d', seq_len=4, token=False, seq_last=False)(t).shape, (2, 3))
test_eq(PoolingLayer('max', token=False, seq_last=False)(t), tensor([[ 3.,  7., 11.], [15., 19., 23.]]))
test_close(PoolingLayer('mean', token=False, seq_last=False)(t), tensor([[ 1.5000,  5.5000,  9.5000],
                                                                        [13.5000, 17.5000, 21.5000]]))
test_close(PoolingLayer('max-mean', token=False, seq_last=False)(t), tensor([[ 3.,  7., 11.,  1.5000,  5.5000,  9.5000],
                                                                            [15., 19., 23., 13.5000, 17.5000, 21.5000]]))
test_close(PoolingLayer('flatten', token=False, seq_last=False)(t), tensor([[ 0.,  4.,  8.,  1.,  5.,  9.,  2.,  6., 10.,  3.,  7., 11.],
                                                                            [12., 16., 20., 13., 17., 21., 14., 18., 22., 15., 19., 23.]]))
test_eq(PoolingLayer('conv1d', seq_len=4, token=False, seq_last=False)(t).shape, (2, 3))

---

source

ReGLU

 ReGLU ()

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

GEGLU

 GEGLU ()

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

get_act_fn

 get_act_fn (act, **act_kwargs)

test_eq(get_act_fn(nn.ReLU).__repr__(), "ReLU()")
test_eq(get_act_fn(nn.ReLU()).__repr__(), "ReLU()")
test_eq(get_act_fn(nn.LeakyReLU, negative_slope=0.05).__repr__(), "LeakyReLU(negative_slope=0.05)")
test_eq(get_act_fn('reglu').__repr__(), "ReGLU()")
test_eq(get_act_fn('leakyrelu', negative_slope=0.05).__repr__(), "LeakyReLU(negative_slope=0.05)")

---

source

RevIN

 RevIN (c_in:int, affine:bool=True, subtract_last:bool=False, dim:int=2,
        eps:float=1e-05)

Reversible Instance Normalization layer adapted from

Kim, T., Kim, J., Tae, Y., Park, C., Choi, J. H., & Choo, J. (2021, September). Reversible instance normalization for accurate time-series forecasting against distribution shift. In International Conference on Learning Representations. Original code: https://github.com/ts-kim/RevIN

	Type	Default	Details
c_in	int		#features (aka variables or channels)
affine	bool	True	flag to incidate if RevIN has learnable weight and bias
subtract_last	bool	False
dim	int	2	int or tuple of dimensions used to calculate mean and std
eps	float	1e-05	epsilon - parameter added for numerical stability

---

source

RevIN

 RevIN (c_in:int, affine:bool=True, subtract_last:bool=False, dim:int=2,
        eps:float=1e-05)

Reversible Instance Normalization layer adapted from

Kim, T., Kim, J., Tae, Y., Park, C., Choi, J. H., & Choo, J. (2021, September). Reversible instance normalization for accurate time-series forecasting against distribution shift. In International Conference on Learning Representations. Original code: https://github.com/ts-kim/RevIN

	Type	Default	Details
c_in	int		#features (aka variables or channels)
affine	bool	True	flag to incidate if RevIN has learnable weight and bias
subtract_last	bool	False
dim	int	2	int or tuple of dimensions used to calculate mean and std
eps	float	1e-05	epsilon - parameter added for numerical stability

t = ((torch.rand(16, 5, 100) - .25) * torch.Tensor([.01, .1, 1, 10, 100]).reshape(1, -1, 1)).cumsum(-1)
t_clone = t.clone()
l = RevIN(5)
t_norm = l(t, torch.tensor(True))
t_denorm = l(t_norm, torch.tensor(False))
test_close(t_clone, t_denorm, eps=1e-3)

model = RevIN(5, affine=True)
try:
    scripted_model = torch.jit.script(model)
    file_path = f"test_scripted_model.pt"
    torch.jit.save(scripted_model, file_path)
    scripted_model = torch.jit.load(file_path)

    inp = ((torch.rand(16, 5, 100) - .25) * torch.Tensor([.01, .1, 1, 10, 100]).reshape(1, -1, 1)).cumsum(-1)
    normed_output = model(inp, torch.tensor(True))
    demormed_output = model(normed_output, torch.tensor(False))
    scripted_normed_output = scripted_model(inp, torch.tensor(True))
    scripted_denormed_output = scripted_model(scripted_normed_output, torch.tensor(False))
    test_close(normed_output, scripted_normed_output)
    test_close(demormed_output, scripted_denormed_output)
    os.remove(file_path)
    del scripted_model
    gc.collect()
    print('scripting ok')
except Exception as e:
    print(f'scripting failed: {e}')

scripting ok

---

source

create_pool_head

 create_pool_head (n_in, c_out, seq_len=None, concat_pool=False,
                   fc_dropout=0.0, bn=False, y_range=None, **kwargs)

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(create_pool_head(nf, c_out, seq_len, fc_dropout=0.5)(t).shape, (bs, c_out))
test_eq(create_pool_head(nf, c_out, seq_len, concat_pool=True, fc_dropout=0.5)(t).shape, (bs, c_out))
create_pool_head(nf, c_out, seq_len, concat_pool=True, bn=True, fc_dropout=.5)

Sequential(
  (0): GACP1d(
    (gacp): AdaptiveConcatPool1d(
      (ap): AdaptiveAvgPool1d(output_size=1)
      (mp): AdaptiveMaxPool1d(output_size=1)
    )
    (flatten): Reshape(bs)
  )
  (1): LinBnDrop(
    (0): BatchNorm1d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (1): Dropout(p=0.5, inplace=False)
    (2): Linear(in_features=24, out_features=2, bias=False)
  )
)

---

source

max_pool_head

 max_pool_head (n_in, c_out, seq_len, fc_dropout=0.0, bn=False,
                y_range=None, **kwargs)

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(max_pool_head(nf, c_out, seq_len, fc_dropout=0.5)(t).shape, (bs, c_out))

---

source

create_pool_plus_head

 create_pool_plus_head (*args, lin_ftrs=None, fc_dropout=0.0,
                        concat_pool=True, bn_final=False, lin_first=False,
                        y_range=None)

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(create_pool_plus_head(nf, c_out, seq_len, fc_dropout=0.5)(t).shape, (bs, c_out))
test_eq(create_pool_plus_head(nf, c_out, concat_pool=True, fc_dropout=0.5)(t).shape, (bs, c_out))
create_pool_plus_head(nf, c_out, seq_len, fc_dropout=0.5)

Sequential(
  (0): AdaptiveConcatPool1d(
    (ap): AdaptiveAvgPool1d(output_size=1)
    (mp): AdaptiveMaxPool1d(output_size=1)
  )
  (1): Reshape(bs)
  (2): BatchNorm1d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (3): Dropout(p=0.25, inplace=False)
  (4): Linear(in_features=24, out_features=512, bias=False)
  (5): ReLU(inplace=True)
  (6): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (7): Dropout(p=0.5, inplace=False)
  (8): Linear(in_features=512, out_features=2, bias=False)
)

---

source

create_conv_head

 create_conv_head (*args, adaptive_size=None, y_range=None)

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(create_conv_head(nf, c_out, seq_len)(t).shape, (bs, c_out))
test_eq(create_conv_head(nf, c_out, adaptive_size=50)(t).shape, (bs, c_out))
create_conv_head(nf, c_out, 50)

Sequential(
  (0): ConvBlock(
    (0): Conv1d(12, 6, kernel_size=(1,), stride=(1,), bias=False)
    (1): BatchNorm1d(6, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (1): ConvBlock(
    (0): Conv1d(6, 3, kernel_size=(1,), stride=(1,), bias=False)
    (1): BatchNorm1d(3, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (2): ConvBlock(
    (0): Conv1d(3, 2, kernel_size=(1,), stride=(1,), bias=False)
    (1): BatchNorm1d(2, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (3): GAP1d(
    (gap): AdaptiveAvgPool1d(output_size=1)
    (flatten): Reshape(bs)
  )
)

---

source

create_mlp_head

 create_mlp_head (nf, c_out, seq_len=None, flatten=True, fc_dropout=0.0,
                  bn=False, lin_first=False, y_range=None)

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(create_mlp_head(nf, c_out, seq_len, fc_dropout=0.5)(t).shape, (bs, c_out))
t = torch.rand(bs, nf, seq_len)
create_mlp_head(nf, c_out, seq_len, bn=True, fc_dropout=.5)

Sequential(
  (0): Reshape(bs)
  (1): LinBnDrop(
    (0): BatchNorm1d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (1): Dropout(p=0.5, inplace=False)
    (2): Linear(in_features=240, out_features=2, bias=False)
  )
)

---

source

create_fc_head

 create_fc_head (nf, c_out, seq_len=None, flatten=True, lin_ftrs=None,
                 y_range=None, fc_dropout=0.0, bn=False, bn_final=False,
                 act=ReLU(inplace=True))

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(create_fc_head(nf, c_out, seq_len, fc_dropout=0.5)(t).shape, (bs, c_out))
create_mlp_head(nf, c_out, seq_len, bn=True, fc_dropout=.5)

Sequential(
  (0): Reshape(bs)
  (1): LinBnDrop(
    (0): BatchNorm1d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (1): Dropout(p=0.5, inplace=False)
    (2): Linear(in_features=240, out_features=2, bias=False)
  )
)

---

source

create_rnn_head

 create_rnn_head (*args, fc_dropout=0.0, bn=False, y_range=None)

bs = 16
nf = 12
c_out = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
test_eq(create_rnn_head(nf, c_out, seq_len, fc_dropout=0.5)(t).shape, (bs, c_out))
create_rnn_head(nf, c_out, seq_len, bn=True, fc_dropout=.5)

Sequential(
  (0): LastStep()
  (1): LinBnDrop(
    (0): BatchNorm1d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (1): Dropout(p=0.5, inplace=False)
    (2): Linear(in_features=12, out_features=2, bias=False)
  )
)

---

source

imputation_head

 imputation_head (c_in, c_out, seq_len=None, ks=1, y_range=None,
                  fc_dropout=0.0)

bs = 16
nf = 12
ni = 2
seq_len = 20
t = torch.rand(bs, nf, seq_len)
head = imputation_head(nf, ni, seq_len=None, ks=1, y_range=None, fc_dropout=0.)
test_eq(head(t).shape, (bs, ni, seq_len))
head = imputation_head(nf, ni, seq_len=None, ks=1, y_range=(.3,.7), fc_dropout=0.)
test_ge(head(t).min(), .3)
test_le(head(t).max(), .7)
y_range = (tensor([0.1000, 0.1000, 0.1000, 0.1000, 0.2000, 0.2000, 0.2000, 0.2000, 0.3000,
                   0.3000, 0.3000, 0.3000]),
           tensor([0.6000, 0.6000, 0.6000, 0.6000, 0.7000, 0.7000, 0.7000, 0.7000, 0.8000,
                   0.8000, 0.8000, 0.8000]))
test_ge(head(t).min(), .1)
test_le(head(t).max(), .9)
head = imputation_head(nf, ni, seq_len=None, ks=1, y_range=y_range, fc_dropout=0.)
head

Sequential(
  (0): Dropout(p=0.0, inplace=False)
  (1): Conv1d(12, 2, kernel_size=(1,), stride=(1,))
  (2): fastai.layers.SigmoidRange(low=tensor([0.1000, 0.1000, 0.1000, 0.1000, 0.2000, 0.2000, 0.2000, 0.2000, 0.3000,
          0.3000, 0.3000, 0.3000]), high=tensor([0.6000, 0.6000, 0.6000, 0.6000, 0.7000, 0.7000, 0.7000, 0.7000, 0.8000,
          0.8000, 0.8000, 0.8000]))
)

---

source

create_conv_lin_nd_head

 create_conv_lin_nd_head (n_in, n_out, seq_len, d, conv_first=True,
                          conv_bn=False, lin_bn=False, fc_dropout=0.0,
                          **kwargs)

Module to create a nd output head

bs = 16
nf = 32
c = 5
seq_len = 10
d = 2
targ = torch.randint(0, c, (bs,d))
t = torch.randn(bs, nf, seq_len)
head = conv_lin_nd_head(nf, c, seq_len, d, conv_first=True, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, (bs, d, c))
loss = CrossEntropyLossFlat()(inp, targ)
loss, head

(TensorBase(1.7074, grad_fn=<AliasBackward0>),
 create_conv_lin_nd_head(
   (0): Conv1d(32, 5, kernel_size=(1,), stride=(1,))
   (1): Dropout(p=0.5, inplace=False)
   (2): Linear(in_features=10, out_features=2, bias=True)
   (3): Transpose(-1, -2)
   (4): Reshape(bs, 2, 5)
 ))

bs = 16
nf = 32
c = 5
seq_len = 10
d = [2, 8]
targ = torch.randint(0, c, [bs]+d)
t = torch.randn(bs, nf, seq_len)
head = conv_lin_nd_head(nf, c, seq_len, d, conv_first=False, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, [bs]+d+[c])
loss = CrossEntropyLossFlat()(inp, targ)
loss, head

(TensorBase(1.6561, grad_fn=<AliasBackward0>),
 create_conv_lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Linear(in_features=10, out_features=16, bias=True)
   (2): Conv1d(32, 5, kernel_size=(1,), stride=(1,))
   (3): Transpose(-1, -2)
   (4): Reshape(bs, 2, 8, 5)
 ))

bs = 16
nf = 32
c = 1
seq_len = 10
d = 2
targ = torch.rand(bs, d)
t = torch.randn(bs, nf, seq_len)
head = conv_lin_nd_head(nf, c, seq_len, d, conv_first=False, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, (bs, d))
loss = L1LossFlat()(inp, targ)
loss, head

(TensorBase(0.6017, grad_fn=<AliasBackward0>),
 create_conv_lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Linear(in_features=10, out_features=2, bias=True)
   (2): Conv1d(32, 1, kernel_size=(1,), stride=(1,))
   (3): Transpose(-1, -2)
   (4): Reshape(bs, 2)
 ))

bs = 16
nf = 32
c = 1
seq_len = 10
d = [2,3]
targ = torch.rand(bs, *d)
t = torch.randn(bs, nf, seq_len)
head = conv_lin_nd_head(nf, c, seq_len, d, conv_first=False, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, [bs]+d)
loss = L1LossFlat()(inp, targ)
loss, head

(TensorBase(0.5439, grad_fn=<AliasBackward0>),
 create_conv_lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Linear(in_features=10, out_features=6, bias=True)
   (2): Conv1d(32, 1, kernel_size=(1,), stride=(1,))
   (3): Transpose(-1, -2)
   (4): Reshape(bs, 2, 3)
 ))

---

source

lin_nd_head

 lin_nd_head (n_in, n_out, seq_len=None, d=None, flatten=False,
              use_bn=False, fc_dropout=0.0)

Module to create a nd output head with linear layers

bs = 16
nf = 32
seq_len = 50
x = torch.normal(0, 1, (bs, nf, seq_len))

for use_bn in [False, True]:
    for fc_dropout in [0, 0.2]:
        for flatten in [False, True]:
            for c in [1, 3]:
                for d in [None, (50,), (50,10), (30,5), (50,2,3), (30,2,3)]:
                    for q_len in [1, seq_len]:
                        head = lin_nd_head(nf, c, q_len, d, flatten=flatten, use_bn=use_bn, fc_dropout=fc_dropout)
                        test_eq(head(x).shape, (bs, ) + (d if d is not None else ()) + ((c,) if c != 1 else ()))

bs = 16
nf = 32
c = 5
seq_len = 10
d = 2
targ = torch.randint(0, c, (bs,d))
t = torch.randn(bs, nf, seq_len)
head = lin_nd_head(nf, c, seq_len, d, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, (bs, d, c))
loss = CrossEntropyLossFlat()(inp, targ)
loss, head

(TensorBase(1.8360, grad_fn=<AliasBackward0>),
 lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Reshape(bs)
   (2): Linear(in_features=320, out_features=10, bias=True)
   (3): Reshape(bs, 2, 5)
 ))

bs = 16
nf = 32
c = 5
seq_len = 10
d = [2, 8]
targ = torch.randint(0, c, [bs]+d)
t = torch.randn(bs, nf, seq_len)
head = lin_nd_head(nf, c, seq_len, d, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, [bs]+d+[c])
loss = CrossEntropyLossFlat()(inp, targ)
loss, head

(TensorBase(1.7557, grad_fn=<AliasBackward0>),
 lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Reshape(bs)
   (2): Linear(in_features=320, out_features=80, bias=True)
   (3): Reshape(bs, 2, 8, 5)
 ))

bs = 16
nf = 32
c = 1
seq_len = 10
d = 2
targ = torch.rand(bs, d)
t = torch.randn(bs, nf, seq_len)
head = lin_nd_head(nf, c, seq_len, d, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, (bs, d))
loss = L1LossFlat()(inp, targ)
loss, head

(TensorBase(0.5978, grad_fn=<AliasBackward0>),
 lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Reshape(bs)
   (2): Linear(in_features=320, out_features=2, bias=True)
   (3): Reshape(bs, 2)
 ))

bs = 16
nf = 32
c = 1
seq_len = 10
d = [2,3]
targ = torch.rand(bs, *d)
t = torch.randn(bs, nf, seq_len)
head = lin_nd_head(nf, c, seq_len, d, fc_dropout=.5)
inp = head(t)
test_eq(inp.shape, [bs]+d)
loss = L1LossFlat()(inp, targ)
loss, head

(TensorBase(0.8286, grad_fn=<AliasBackward0>),
 lin_nd_head(
   (0): Dropout(p=0.5, inplace=False)
   (1): Reshape(bs)
   (2): Linear(in_features=320, out_features=6, bias=True)
   (3): Reshape(bs, 2, 3)
 ))

---

source

rocket_nd_head

 rocket_nd_head (n_in, n_out, seq_len=None, d=None, use_bn=False,
                 fc_dropout=0.0, zero_init=True)

Module to create a nd output head with linear layers for the rocket family of models

bs = 16
nf = 99
seq_len = 1
x = torch.normal(0, 1, (bs, nf, seq_len))

for use_bn in [False, True]:
    for fc_dropout in [0, 0.2]:
        for c in [1, 3]:
            for d in [None, (50,), (50,10), (30,5), (50,2,3), (30,2,3)]:
                head = rocket_nd_head(nf, c, 1, d, use_bn=use_bn, fc_dropout=fc_dropout)
                test_eq(head(x).shape, (bs, ) + (d if d is not None else ()) + ((c,) if c != 1 else ()))

---

source

xresnet1d_nd_head

 xresnet1d_nd_head (n_in, n_out, seq_len=None, d=None, use_bn=False,
                    fc_dropout=0.0, zero_init=True)

Module to create a nd output head with linear layers for the xresnet family of models

bs = 16
nf = 99
seq_len = 2
x = torch.normal(0, 1, (bs, nf, seq_len))

for use_bn in [False, True]:
    for fc_dropout in [0, 0.2]:
        for c in [1, 3]:
            for d in [None, (50,), (50,10), (30,5), (50,2,3), (30,2,3)]:
                head = xresnet1d_nd_head(nf, c, 1, d, use_bn=use_bn, fc_dropout=fc_dropout)
                test_eq(head(x).shape, (bs, ) + (d if d is not None else ()) + ((c,) if c != 1 else ()))

---

source

create_conv_3d_head

 create_conv_3d_head (n_in, n_out, seq_len, d, use_bn=False, **kwargs)

Module to create a nd output head with a convolutional layer

bs = 16
nf = 32
c = 5
seq_len = 10
d = 10
targ = torch.randint(0, c, (bs,d))
t = torch.randn(bs, nf, seq_len)
head = conv_3d_head(nf, c, seq_len, d)
inp = head(t)
test_eq(inp.shape, (bs, d, c))
loss = CrossEntropyLossFlat()(inp, targ)
loss, head

(TensorBase(1.7321, grad_fn=<AliasBackward0>),
 create_conv_3d_head(
   (0): ConvBlock(
     (0): Conv1d(32, 5, kernel_size=(1,), stride=(1,))
   )
   (1): Transpose(-1, -2)
 ))

bs = 16
nf = 32
c = 1
seq_len = 10
d = 10
targ = torch.rand(bs, d)
t = torch.randn(bs, nf, seq_len)
head = conv_3d_head(nf, c, seq_len, d)
inp = head(t)
test_eq(inp.shape, (bs, d))
loss = L1LossFlat()(inp, targ)
loss, head

(TensorBase(0.5833, grad_fn=<AliasBackward0>),
 create_conv_3d_head(
   (0): ConvBlock(
     (0): Conv1d(32, 1, kernel_size=(1,), stride=(1,))
   )
   (1): Transpose(-1, -2)
   (2): Squeeze(dim=-1)
 ))

---

source

universal_pool_head

 universal_pool_head (n_in, c_out, seq_len, mult=2, pool_n_layers=2,
                      pool_ln=True, pool_dropout=0.5, pool_act=ReLU(),
                      zero_init=True, bn=True, fc_dropout=0.0)

bs, c_in, seq_len = 16, 128, 50
c_out = 14
t = torch.rand(bs, c_in, seq_len)
uph = universal_pool_head(c_in, c_out, seq_len)
test_eq(uph(t).shape, (bs, c_out))
uph = universal_pool_head(c_in, c_out, seq_len, 2)
test_eq(uph(t).shape, (bs, c_out))

bs, c_in, seq_len = 16, 128, 50
c_out = 14
d = 5
t = torch.rand(bs, c_in, seq_len)
for head in heads: 
    print(head.__name__)
    if head.__name__ == "create_conv_3d_head":
        h = head(c_in, c_out, seq_len, seq_len)
        test_eq(h(t).shape, (bs, seq_len, c_out))
    elif 'nd' in head.__name__: 
        h = head(c_in, c_out, seq_len, d)
        test_eq(h(t).shape, (bs, d, c_out))
    else: 
        h = head(c_in, c_out, seq_len)
        test_eq(h(t).shape, (bs, c_out))

create_mlp_head
create_fc_head
average_pool_head
max_pool_head
concat_pool_head
create_pool_plus_head
create_conv_head
create_rnn_head
create_conv_lin_nd_head
lin_nd_head
create_conv_3d_head
attentional_pool_head
universal_pool_head
gwa_pool_head

---

source

SqueezeExciteBlock

 SqueezeExciteBlock (ni, reduction=16)

Same as nn.Module, but no need for subclasses to call super().__init__

bs = 2
ni = 32
sl = 4
t = torch.rand(bs, ni, sl)
test_eq(SqueezeExciteBlock(ni)(t).shape, (bs, ni, sl))

---

source

GaussianNoise

 GaussianNoise (sigma=0.1, is_relative_detach=True)

Gaussian noise regularizer.

Args: sigma (float, optional): relative standard deviation used to generate the noise. Relative means that it will be multiplied by the magnitude of the value your are adding the noise to. This means that sigma can be the same regardless of the scale of the vector. is_relative_detach (bool, optional): whether to detach the variable before computing the scale of the noise. If False then the scale of the noise won’t be seen as a constant but something to optimize: this will bias the network to generate vectors with smaller values.

t = torch.ones(2,3,4)
test_ne(GaussianNoise()(t), t)
test_eq(GaussianNoise()(t).shape, t.shape)
t = torch.ones(2,3)
test_ne(GaussianNoise()(t), t)
test_eq(GaussianNoise()(t).shape, t.shape)
t = torch.ones(2)
test_ne(GaussianNoise()(t), t)
test_eq(GaussianNoise()(t).shape, t.shape)

---

source

TokenLayer

 TokenLayer (token=True)

Same as nn.Module, but no need for subclasses to call super().__init__

---

source

PositionwiseFeedForward

 PositionwiseFeedForward (dim, dropout=0.0, act='reglu', mlp_ratio=1)

A sequential container.

Modules will be added to it in the order they are passed in the constructor. Alternatively, an OrderedDict of modules can be passed in. The forward() method of [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) accepts any input and forwards it to the first module it contains. It then “chains” outputs to inputs sequentially for each subsequent module, finally returning the output of the last module.

The value a [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) provides over manually calling a sequence of modules is that it allows treating the whole container as a single module, such that performing a transformation on the [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) applies to each of the modules it stores (which are each a registered submodule of the [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential)).

What’s the difference between a [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) and a :class:torch.nn.ModuleList? A ModuleList is exactly what it sounds like–a list for storing Module s! On the other hand, the layers in a [Sequential](https://timeseriesAI.github.io/models.layers.html#sequential) are connected in a cascading way.

Example::

# Using Sequential to create a small model. When `model` is run,
# input will first be passed to `Conv2d(1,20,5)`. The output of
# `Conv2d(1,20,5)` will be used as the input to the first
# `ReLU`; the output of the first `ReLU` will become the input
# for `Conv2d(20,64,5)`. Finally, the output of
# `Conv2d(20,64,5)` will be used as input to the second `ReLU`
model = nn.Sequential(
          nn.Conv2d(1,20,5),
          nn.ReLU(),
          nn.Conv2d(20,64,5),
          nn.ReLU()
        )

# Using Sequential with OrderedDict. This is functionally the
# same as the above code
model = nn.Sequential(OrderedDict([
          ('conv1', nn.Conv2d(1,20,5)),
          ('relu1', nn.ReLU()),
          ('conv2', nn.Conv2d(20,64,5)),
          ('relu2', nn.ReLU())
        ]))

t = torch.randn(2,3,10)
m = PositionwiseFeedForward(10, dropout=0., act='reglu', mlp_ratio=1)
test_eq(m(t).shape, t.shape)
m = PositionwiseFeedForward(10, dropout=0., act='smelu', mlp_ratio=1)
test_eq(m(t).shape, t.shape)

---

source

ScaledDotProductAttention

 ScaledDotProductAttention (d_model, n_heads, attn_dropout=0.0,
                            res_attention=False, lsa=False)

Scaled Dot-Product Attention module (Attention is all you need by Vaswani et al., 2017) with optional residual attention from previous layer (Realformer: Transformer likes residual attention by He et al, 2020) and locality self sttention (Vision Transformer for Small-Size Datasets by Lee et al, 2021)

B = 16
C = 10
M = 1500 # seq_len

n_heads = 1
D = 128 # model dimension
N = 512 # max_seq_len - latent's index dimension
d_k = D // n_heads

xb = torch.randn(B, C, M)
xb = (xb - xb.mean()) / xb.std()

# Attention
# input (Q)
lin = nn.Linear(M, N, bias=False)
Q = lin(xb).transpose(1,2)
test_eq(Q.shape, (B, N, C))

# q
to_q = nn.Linear(C, D, bias=False)
q = to_q(Q)
q = nn.LayerNorm(D)(q)

# k, v
context = xb.transpose(1,2)
to_kv = nn.Linear(C, D * 2, bias=False)
k, v = to_kv(context).chunk(2, dim = -1)
k = k.transpose(-1, -2)
k = nn.LayerNorm(M)(k)
v = nn.LayerNorm(D)(v)

test_eq(q.shape, (B, N, D))
test_eq(k.shape, (B, D, M))
test_eq(v.shape, (B, M, D))

output, attn, scores = ScaledDotProductAttention(D, n_heads, res_attention=True)(q.unsqueeze(1), k.unsqueeze(1), v.unsqueeze(1))
test_eq(output.shape, (B, 1, N, D))
test_eq(attn.shape, (B, 1, N, M))
test_eq(scores.shape, (B, 1, N, M))
scores.mean(), scores.std()

(tensor(1.3535e-10, grad_fn=<MeanBackward0>),
 tensor(1.0555, grad_fn=<StdBackward0>))

---

source

MultiheadAttention

 MultiheadAttention (d_model, n_heads, d_k=None, d_v=None,
                     res_attention=False, attn_dropout=0.0,
                     proj_dropout=0.0, qkv_bias=True, lsa=False)

Same as nn.Module, but no need for subclasses to call super().__init__

q = torch.rand([16, 3, 50, 8]) 
k = torch.rand([16, 3, 50, 8]).transpose(-1, -2)
v = torch.rand([16, 3, 50, 6])
attn_mask = torch.triu(torch.ones(50, 50)) # shape: q_len x q_len
key_padding_mask = torch.zeros(16, 50)
key_padding_mask[[1, 3, 6, 15], -10:] = 1
key_padding_mask = key_padding_mask.bool()
print('attn_mask', attn_mask.shape, 'key_padding_mask', key_padding_mask.shape)
output, attn = ScaledDotProductAttention(24, 3, attn_dropout=.1)(q, k, v, attn_mask=attn_mask, key_padding_mask=key_padding_mask)
output.shape, attn.shape

attn_mask torch.Size([50, 50]) key_padding_mask torch.Size([16, 50])

(torch.Size([16, 3, 50, 6]), torch.Size([16, 3, 50, 50]))

t = torch.rand(16, 50, 128)
output, attn = MultiheadAttention(d_model=128, n_heads=3, d_k=8, d_v=6)(t, t, t, key_padding_mask=key_padding_mask, attn_mask=attn_mask)
output.shape, attn.shape

(torch.Size([16, 50, 128]), torch.Size([16, 3, 50, 50]))

Test multi-head attention with self-locality attention

# lsa (locality self-sttention)
t = torch.rand(16, 50, 128)
attn_mask = torch.eye(50).reshape(1, 1, 50, 50).bool()
output, attn = MultiheadAttention(d_model=128, n_heads=8, lsa=True)(t, t, t, key_padding_mask=key_padding_mask, attn_mask=attn_mask)
output.shape, attn.shape

(torch.Size([16, 50, 128]), torch.Size([16, 8, 50, 50]))

t = torch.rand(16, 50, 128)
att_mask = (torch.rand((50, 50)) > .85).float()
att_mask[att_mask == 1] = -np.inf

mha = MultiheadAttention(d_model=128, n_heads=3, d_k=8, d_v=6)
output, attn = mha(t, t, t, attn_mask=att_mask)
test_eq(torch.isnan(output).sum().item(), 0)
test_eq(torch.isnan(attn).sum().item(), 0)
loss = output[:2, :].sum()
test_eq(torch.isnan(loss).sum().item(), 0)
loss.backward()
for n, p in mha.named_parameters(): 
    if p.grad is not None:
        test_eq(torch.isnan(p.grad).sum().item(), 0)

t = torch.rand(16, 50, 128)
attn_mask = (torch.rand((50, 50)) > .85)

# True values will be masked
mha = MultiheadAttention(d_model=128, n_heads=3, d_k=8, d_v=6)
output, attn = mha(t, t, t, attn_mask=att_mask)
test_eq(torch.isnan(output).sum().item(), 0)
test_eq(torch.isnan(attn).sum().item(), 0)
loss = output[:2, :].sum()
test_eq(torch.isnan(loss).sum().item(), 0)
loss.backward()
for n, p in mha.named_parameters(): 
    if p.grad is not None:
        test_eq(torch.isnan(p.grad).sum().item(), 0)

---

source

MultiConv1d

 MultiConv1d (ni, nf=None, kss=[1, 3, 5, 7], keep_original=False,
              separable=False, dim=1, **kwargs)

Module that applies multiple convolutions with different kernel sizes

t = torch.rand(16, 6, 37)
test_eq(MultiConv1d(6, None, kss=[1,3,5], keep_original=True)(t).shape, [16, 24, 37])
test_eq(MultiConv1d(6, 36, kss=[1,3,5], keep_original=False)(t).shape, [16, 36, 37])
test_eq(MultiConv1d(6, None, kss=[1,3,5], keep_original=True, dim=-1)(t).shape, [16, 6, 37*4])
test_eq(MultiConv1d(6, 60, kss=[1,3,5], keep_original=True)(t).shape, [16, 60, 37])
test_eq(MultiConv1d(6, 60, kss=[1,3,5], separable=True)(t).shape, [16, 60, 37])

---

source

LSTMOutput

 LSTMOutput ()

Same as nn.Module, but no need for subclasses to call super().__init__

t = ([1], [2], [3])
test_eq(LSTMOutput()(t), [1])

---

source

emb_sz_rule

 emb_sz_rule (n_cat)

Rule of thumb to pick embedding size corresponding to n_cat (original from fastai)

test_eq(emb_sz_rule(7), 5)

---

source

TSEmbedding

 TSEmbedding (ni, nf, std=0.01, padding_idx=None)

Embedding layer with truncated normal initialization adapted from fastai

---

source

MultiEmbedding

 MultiEmbedding (c_in, n_cat_embeds, cat_embed_dims=None, cat_pos=None,
                 std=0.01, cat_padding_idxs=None)

Same as nn.Module, but no need for subclasses to call super().__init__

a = alphabet[np.random.randint(0,3,40)]
b = ALPHABET[np.random.randint(6,10,40)]
c = np.random.rand(40).reshape(4,1,10)
map_a = {k:v for v,k in enumerate(np.unique(a))}
map_b = {k:v for v,k in enumerate(np.unique(b))}
n_embeds = [len(m.keys()) for m in [map_a, map_b]]
szs = [emb_sz_rule(n) for n in n_embeds]
a = np.asarray(a.map(map_a)).reshape(4,1,10)
b = np.asarray(b.map(map_b)).reshape(4,1,10)
inp = torch.from_numpy(np.concatenate((c,a,b), 1)).float()
memb = MultiEmbedding(3, n_embeds, cat_pos=[1,2])
# registered buffers are part of the state_dict() but not module.parameters()
assert all([(k in memb.state_dict().keys()) for k in ['cat_pos', 'cont_pos']])
embeddings = memb(inp)
print(n_embeds, szs, inp.shape, embeddings.shape)
test_eq(embeddings.shape, (inp.shape[0],sum(szs)+1,inp.shape[-1]))

[3, 4] [3, 3] torch.Size([4, 3, 10]) torch.Size([4, 7, 10])

me = MultiEmbedding(3, 4, cat_pos=2)
test_eq(me.cat_embed[0].weight.shape, (4,3))
test_eq(me.cat_pos.cpu().item(), 2)