torch/nn/init.py

# mypy: allow-untyped-defs
"""This file contains utilities for initializing neural network parameters."""
import math
import warnings
from typing import Optional as _Optional

import torch
from torch import Tensor


# These no_grad_* functions are necessary as wrappers around the parts of these
# functions that use `with torch.no_grad()`. The JIT doesn't support context
# managers, so these need to be implemented as builtins. Using these wrappers
# lets us keep those builtins small and re-usable.
def _no_grad_uniform_(tensor, a, b, generator=None):
    with torch.no_grad():
        return tensor.uniform_(a, b, generator=generator)


def _no_grad_normal_(tensor, mean, std, generator=None):
    with torch.no_grad():
        return tensor.normal_(mean, std, generator=generator)


def _no_grad_trunc_normal_(tensor, mean, std, a, b, generator=None):
    # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf
    def norm_cdf(x):
        # Computes standard normal cumulative distribution function
        return (1.0 + math.erf(x / math.sqrt(2.0))) / 2.0

    if (mean < a - 2 * std) or (mean > b + 2 * std):
        warnings.warn(
            "mean is more than 2 std from [a, b] in nn.init.trunc_normal_. "
            "The distribution of values may be incorrect.",
            stacklevel=2,
        )

    with torch.no_grad():
        # Values are generated by using a truncated uniform distribution and
        # then using the inverse CDF for the normal distribution.
        # Get upper and lower cdf values
        l = norm_cdf((a - mean) / std)
        u = norm_cdf((b - mean) / std)

        # Uniformly fill tensor with values from [l, u], then translate to
        # [2l-1, 2u-1].
        tensor.uniform_(2 * l - 1, 2 * u - 1, generator=generator)

        # Use inverse cdf transform for normal distribution to get truncated
        # standard normal
        tensor.erfinv_()

        # Transform to proper mean, std
        tensor.mul_(std * math.sqrt(2.0))
        tensor.add_(mean)

        # Clamp to ensure it's in the proper range
        tensor.clamp_(min=a, max=b)
        return tensor


def _no_grad_fill_(tensor, val):
    with torch.no_grad():
        return tensor.fill_(val)


def _no_grad_zero_(tensor):
    with torch.no_grad():
        return tensor.zero_()


def calculate_gain(nonlinearity, param=None):
    r"""Return the recommended gain value for the given nonlinearity function.

    The values are as follows:

    ================= ====================================================
    nonlinearity      gain
    ================= ====================================================
    Linear / Identity :math:`1`
    Conv{1,2,3}D      :math:`1`
    Sigmoid           :math:`1`
    Tanh              :math:`\frac{5}{3}`
    ReLU              :math:`\sqrt{2}`
    Leaky Relu        :math:`\sqrt{\frac{2}{1 + \text{negative\_slope}^2}}`
    SELU              :math:`\frac{3}{4}`
    ================= ====================================================

    .. warning::
        In order to implement `Self-Normalizing Neural Networks`_ ,
        you should use ``nonlinearity='linear'`` instead of ``nonlinearity='selu'``.
        This gives the initial weights a variance of ``1 / N``,
        which is necessary to induce a stable fixed point in the forward pass.
        In contrast, the default gain for ``SELU`` sacrifices the normalization
        effect for more stable gradient flow in rectangular layers.

    Args:
        nonlinearity: the non-linear function (`nn.functional` name)
        param: optional parameter for the non-linear function

    Examples:
        >>> gain = nn.init.calculate_gain('leaky_relu', 0.2)  # leaky_relu with negative_slope=0.2

    .. _Self-Normalizing Neural Networks: https://papers.nips.cc/paper/2017/hash/5d44ee6f2c3f71b73125876103c8f6c4-Abstract.html
    """
    linear_fns = [
        "linear",
        "conv1d",
        "conv2d",
        "conv3d",
        "conv_transpose1d",
        "conv_transpose2d",
        "conv_transpose3d",
    ]
    if nonlinearity in linear_fns or nonlinearity == "sigmoid":
        return 1
    elif nonlinearity == "tanh":
        return 5.0 / 3
    elif nonlinearity == "relu":
        return math.sqrt(2.0)
    elif nonlinearity == "leaky_relu":
        if param is None:
            negative_slope = 0.01
        elif (
            not isinstance(param, bool)
            and isinstance(param, int)
            or isinstance(param, float)
        ):
            # True/False are instances of int, hence check above
            negative_slope = param
        else:
            raise ValueError(f"negative_slope {param} not a valid number")
        return math.sqrt(2.0 / (1 + negative_slope**2))
    elif nonlinearity == "selu":
        return (
            3.0 / 4
        )  # Value found empirically (https://github.com/pytorch/pytorch/pull/50664)
    else:
        raise ValueError(f"Unsupported nonlinearity {nonlinearity}")


def uniform_(
    tensor: Tensor,
    a: float = 0.0,
    b: float = 1.0,
    generator: _Optional[torch.Generator] = None,
) -> Tensor:
    r"""Fill the input Tensor with values drawn from the uniform distribution.

    :math:`\mathcal{U}(a, b)`.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        a: the lower bound of the uniform distribution
        b: the upper bound of the uniform distribution
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.uniform_(w)
    """
    if torch.overrides.has_torch_function_variadic(tensor):
        return torch.overrides.handle_torch_function(
            uniform_, (tensor,), tensor=tensor, a=a, b=b, generator=generator
        )
    return _no_grad_uniform_(tensor, a, b, generator)


def normal_(
    tensor: Tensor,
    mean: float = 0.0,
    std: float = 1.0,
    generator: _Optional[torch.Generator] = None,
) -> Tensor:
    r"""Fill the input Tensor with values drawn from the normal distribution.

    :math:`\mathcal{N}(\text{mean}, \text{std}^2)`.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        mean: the mean of the normal distribution
        std: the standard deviation of the normal distribution
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.normal_(w)
    """
    if torch.overrides.has_torch_function_variadic(tensor):
        return torch.overrides.handle_torch_function(
            normal_, (tensor,), tensor=tensor, mean=mean, std=std, generator=generator
        )
    return _no_grad_normal_(tensor, mean, std, generator)


def trunc_normal_(
    tensor: Tensor,
    mean: float = 0.0,
    std: float = 1.0,
    a: float = -2.0,
    b: float = 2.0,
    generator: _Optional[torch.Generator] = None,
) -> Tensor:
    r"""Fill the input Tensor with values drawn from a truncated normal distribution.

    The values are effectively drawn from the
    normal distribution :math:`\mathcal{N}(\text{mean}, \text{std}^2)`
    with values outside :math:`[a, b]` redrawn until they are within
    the bounds. The method used for generating the random values works
    best when :math:`a \leq \text{mean} \leq b`.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        mean: the mean of the normal distribution
        std: the standard deviation of the normal distribution
        a: the minimum cutoff value
        b: the maximum cutoff value
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.trunc_normal_(w)
    """
    return _no_grad_trunc_normal_(tensor, mean, std, a, b, generator=generator)


def constant_(tensor: Tensor, val: float) -> Tensor:
    r"""Fill the input Tensor with the value :math:`\text{val}`.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        val: the value to fill the tensor with

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.constant_(w, 0.3)
    """
    if torch.overrides.has_torch_function_variadic(tensor):
        return torch.overrides.handle_torch_function(
            constant_, (tensor,), tensor=tensor, val=val
        )
    return _no_grad_fill_(tensor, val)


def ones_(tensor: Tensor) -> Tensor:
    r"""Fill the input Tensor with the scalar value `1`.

    Args:
        tensor: an n-dimensional `torch.Tensor`

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.ones_(w)
    """
    return _no_grad_fill_(tensor, 1.0)


def zeros_(tensor: Tensor) -> Tensor:
    r"""Fill the input Tensor with the scalar value `0`.

    Args:
        tensor: an n-dimensional `torch.Tensor`

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.zeros_(w)
    """
    return _no_grad_zero_(tensor)


def eye_(tensor):
    r"""Fill the 2-dimensional input `Tensor` with the identity matrix.

    Preserves the identity of the inputs in `Linear` layers, where as
    many inputs are preserved as possible.

    Args:
        tensor: a 2-dimensional `torch.Tensor`

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.eye_(w)
    """
    if tensor.ndimension() != 2:
        raise ValueError("Only tensors with 2 dimensions are supported")

    with torch.no_grad():
        torch.eye(*tensor.shape, out=tensor, requires_grad=tensor.requires_grad)
    return tensor


def dirac_(tensor, groups=1):
    r"""Fill the {3, 4, 5}-dimensional input `Tensor` with the Dirac delta function.

    Preserves the identity of the inputs in `Convolutional`
    layers, where as many input channels are preserved as possible. In case
    of groups>1, each group of channels preserves identity

    Args:
        tensor: a {3, 4, 5}-dimensional `torch.Tensor`
        groups (int, optional): number of groups in the conv layer (default: 1)
    Examples:
        >>> w = torch.empty(3, 16, 5, 5)
        >>> nn.init.dirac_(w)
        >>> w = torch.empty(3, 24, 5, 5)
        >>> nn.init.dirac_(w, 3)
    """
    dimensions = tensor.ndimension()
    if dimensions not in [3, 4, 5]:
        raise ValueError("Only tensors with 3, 4, or 5 dimensions are supported")

    sizes = tensor.size()

    if sizes[0] % groups != 0:
        raise ValueError("dim 0 must be divisible by groups")

    out_chans_per_grp = sizes[0] // groups
    min_dim = min(out_chans_per_grp, sizes[1])

    with torch.no_grad():
        tensor.zero_()

        for g in range(groups):
            for d in range(min_dim):
                if dimensions == 3:  # Temporal convolution
                    tensor[g * out_chans_per_grp + d, d, tensor.size(2) // 2] = 1
                elif dimensions == 4:  # Spatial convolution
                    tensor[
                        g * out_chans_per_grp + d,
                        d,
                        tensor.size(2) // 2,
                        tensor.size(3) // 2,
                    ] = 1
                else:  # Volumetric convolution
                    tensor[
                        g * out_chans_per_grp + d,
                        d,
                        tensor.size(2) // 2,
                        tensor.size(3) // 2,
                        tensor.size(4) // 2,
                    ] = 1
    return tensor


def _calculate_fan_in_and_fan_out(tensor):
    dimensions = tensor.dim()
    if dimensions < 2:
        raise ValueError(
            "Fan in and fan out can not be computed for tensor with fewer than 2 dimensions"
        )

    num_input_fmaps = tensor.size(1)
    num_output_fmaps = tensor.size(0)
    receptive_field_size = 1
    if tensor.dim() > 2:
        # math.prod is not always available, accumulate the product manually
        # we could use functools.reduce but that is not supported by TorchScript
        for s in tensor.shape[2:]:
            receptive_field_size *= s
    fan_in = num_input_fmaps * receptive_field_size
    fan_out = num_output_fmaps * receptive_field_size

    return fan_in, fan_out


def xavier_uniform_(
    tensor: Tensor,
    gain: float = 1.0,
    generator: _Optional[torch.Generator] = None,
) -> Tensor:
    r"""Fill the input `Tensor` with values using a Xavier uniform distribution.

    The method is described in `Understanding the difficulty of training
    deep feedforward neural networks` - Glorot, X. & Bengio, Y. (2010).
    The resulting tensor will have values sampled from
    :math:`\mathcal{U}(-a, a)` where

    .. math::
        a = \text{gain} \times \sqrt{\frac{6}{\text{fan\_in} + \text{fan\_out}}}

    Also known as Glorot initialization.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        gain: an optional scaling factor
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.xavier_uniform_(w, gain=nn.init.calculate_gain('relu'))

    Note:
        Be aware that ``fan_in`` and ``fan_out`` are calculated assuming
        that the weight matrix is used in a transposed manner,
        (i.e., ``x @ w.T`` in ``Linear`` layers, where ``w.shape = [fan_out, fan_in]``).
        This is important for correct initialization.
        If you plan to use ``x @ w``, where ``w.shape = [fan_in, fan_out]``,
        pass in a transposed weight matrix, i.e. ``nn.init.xavier_uniform_(w.T, ...)``.
    """
    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor)
    std = gain * math.sqrt(2.0 / float(fan_in + fan_out))
    a = math.sqrt(3.0) * std  # Calculate uniform bounds from standard deviation

    return _no_grad_uniform_(tensor, -a, a, generator)


def xavier_normal_(
    tensor: Tensor,
    gain: float = 1.0,
    generator: _Optional[torch.Generator] = None,
) -> Tensor:
    r"""Fill the input `Tensor` with values using a Xavier normal distribution.

    The method is described in `Understanding the difficulty of training deep feedforward
    neural networks` - Glorot, X. & Bengio, Y. (2010). The resulting tensor
    will have values sampled from :math:`\mathcal{N}(0, \text{std}^2)` where

    .. math::
        \text{std} = \text{gain} \times \sqrt{\frac{2}{\text{fan\_in} + \text{fan\_out}}}

    Also known as Glorot initialization.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        gain: an optional scaling factor
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.xavier_normal_(w)

    Note:
        Be aware that ``fan_in`` and ``fan_out`` are calculated assuming
        that the weight matrix is used in a transposed manner,
        (i.e., ``x @ w.T`` in ``Linear`` layers, where ``w.shape = [fan_out, fan_in]``).
        This is important for correct initialization.
        If you plan to use ``x @ w``, where ``w.shape = [fan_in, fan_out]``,
        pass in a transposed weight matrix, i.e. ``nn.init.xavier_normal_(w.T, ...)``.
    """
    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor)
    std = gain * math.sqrt(2.0 / float(fan_in + fan_out))

    return _no_grad_normal_(tensor, 0.0, std, generator)


def _calculate_correct_fan(tensor, mode):
    mode = mode.lower()
    valid_modes = ["fan_in", "fan_out"]
    if mode not in valid_modes:
        raise ValueError(f"Mode {mode} not supported, please use one of {valid_modes}")

    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor)
    return fan_in if mode == "fan_in" else fan_out


def kaiming_uniform_(
    tensor: Tensor,
    a: float = 0,
    mode: str = "fan_in",
    nonlinearity: str = "leaky_relu",
    generator: _Optional[torch.Generator] = None,
):
    r"""Fill the input `Tensor` with values using a Kaiming uniform distribution.

    The method is described in `Delving deep into rectifiers: Surpassing
    human-level performance on ImageNet classification` - He, K. et al. (2015).
    The resulting tensor will have values sampled from
    :math:`\mathcal{U}(-\text{bound}, \text{bound})` where

    .. math::
        \text{bound} = \text{gain} \times \sqrt{\frac{3}{\text{fan\_mode}}}

    Also known as He initialization.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        a: the negative slope of the rectifier used after this layer (only
            used with ``'leaky_relu'``)
        mode: either ``'fan_in'`` (default) or ``'fan_out'``. Choosing ``'fan_in'``
            preserves the magnitude of the variance of the weights in the
            forward pass. Choosing ``'fan_out'`` preserves the magnitudes in the
            backwards pass.
        nonlinearity: the non-linear function (`nn.functional` name),
            recommended to use only with ``'relu'`` or ``'leaky_relu'`` (default).
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.kaiming_uniform_(w, mode='fan_in', nonlinearity='relu')

    Note:
        Be aware that ``fan_in`` and ``fan_out`` are calculated assuming
        that the weight matrix is used in a transposed manner,
        (i.e., ``x @ w.T`` in ``Linear`` layers, where ``w.shape = [fan_out, fan_in]``).
        This is important for correct initialization.
        If you plan to use ``x @ w``, where ``w.shape = [fan_in, fan_out]``,
        pass in a transposed weight matrix, i.e. ``nn.init.kaiming_uniform_(w.T, ...)``.
    """
    if torch.overrides.has_torch_function_variadic(tensor):
        return torch.overrides.handle_torch_function(
            kaiming_uniform_,
            (tensor,),
            tensor=tensor,
            a=a,
            mode=mode,
            nonlinearity=nonlinearity,
            generator=generator,
        )

    if 0 in tensor.shape:
        warnings.warn("Initializing zero-element tensors is a no-op")
        return tensor
    fan = _calculate_correct_fan(tensor, mode)
    gain = calculate_gain(nonlinearity, a)
    std = gain / math.sqrt(fan)
    bound = math.sqrt(3.0) * std  # Calculate uniform bounds from standard deviation
    with torch.no_grad():
        return tensor.uniform_(-bound, bound, generator=generator)


def kaiming_normal_(
    tensor: Tensor,
    a: float = 0,
    mode: str = "fan_in",
    nonlinearity: str = "leaky_relu",
    generator: _Optional[torch.Generator] = None,
):
    r"""Fill the input `Tensor` with values using a Kaiming normal distribution.

    The method is described in `Delving deep into rectifiers: Surpassing
    human-level performance on ImageNet classification` - He, K. et al. (2015).
    The resulting tensor will have values sampled from
    :math:`\mathcal{N}(0, \text{std}^2)` where

    .. math::
        \text{std} = \frac{\text{gain}}{\sqrt{\text{fan\_mode}}}

    Also known as He initialization.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        a: the negative slope of the rectifier used after this layer (only
            used with ``'leaky_relu'``)
        mode: either ``'fan_in'`` (default) or ``'fan_out'``. Choosing ``'fan_in'``
            preserves the magnitude of the variance of the weights in the
            forward pass. Choosing ``'fan_out'`` preserves the magnitudes in the
            backwards pass.
        nonlinearity: the non-linear function (`nn.functional` name),
            recommended to use only with ``'relu'`` or ``'leaky_relu'`` (default).
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.kaiming_normal_(w, mode='fan_out', nonlinearity='relu')

    Note:
        Be aware that ``fan_in`` and ``fan_out`` are calculated assuming
        that the weight matrix is used in a transposed manner,
        (i.e., ``x @ w.T`` in ``Linear`` layers, where ``w.shape = [fan_out, fan_in]``).
        This is important for correct initialization.
        If you plan to use ``x @ w``, where ``w.shape = [fan_in, fan_out]``,
        pass in a transposed weight matrix, i.e. ``nn.init.kaiming_normal_(w.T, ...)``.
    """
    if 0 in tensor.shape:
        warnings.warn("Initializing zero-element tensors is a no-op")
        return tensor
    fan = _calculate_correct_fan(tensor, mode)
    gain = calculate_gain(nonlinearity, a)
    std = gain / math.sqrt(fan)
    with torch.no_grad():
        return tensor.normal_(0, std, generator=generator)


def orthogonal_(
    tensor,
    gain=1,
    generator: _Optional[torch.Generator] = None,
):
    r"""Fill the input `Tensor` with a (semi) orthogonal matrix.

    Described in `Exact solutions to the nonlinear dynamics of learning in deep
    linear neural networks` - Saxe, A. et al. (2013). The input tensor must have
    at least 2 dimensions, and for tensors with more than 2 dimensions the
    trailing dimensions are flattened.

    Args:
        tensor: an n-dimensional `torch.Tensor`, where :math:`n \geq 2`
        gain: optional scaling factor
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_LAPACK)
        >>> w = torch.empty(3, 5)
        >>> nn.init.orthogonal_(w)
    """
    if tensor.ndimension() < 2:
        raise ValueError("Only tensors with 2 or more dimensions are supported")

    if tensor.numel() == 0:
        # no-op
        return tensor
    rows = tensor.size(0)
    cols = tensor.numel() // rows
    flattened = tensor.new_empty((rows, cols)).normal_(0, 1, generator=generator)

    if rows < cols:
        flattened.t_()

    # Compute the qr factorization
    q, r = torch.linalg.qr(flattened)
    # Make Q uniform according to https://arxiv.org/pdf/math-ph/0609050.pdf
    d = torch.diag(r, 0)
    ph = d.sign()
    q *= ph

    if rows < cols:
        q.t_()

    with torch.no_grad():
        tensor.view_as(q).copy_(q)
        tensor.mul_(gain)
    return tensor


def sparse_(
    tensor,
    sparsity,
    std=0.01,
    generator: _Optional[torch.Generator] = None,
):
    r"""Fill the 2D input `Tensor` as a sparse matrix.

    The non-zero elements will be drawn from the normal distribution
    :math:`\mathcal{N}(0, 0.01)`, as described in `Deep learning via
    Hessian-free optimization` - Martens, J. (2010).

    Args:
        tensor: an n-dimensional `torch.Tensor`
        sparsity: The fraction of elements in each column to be set to zero
        std: the standard deviation of the normal distribution used to generate
            the non-zero values
        generator: the torch Generator to sample from (default: None)

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.sparse_(w, sparsity=0.1)
    """
    if tensor.ndimension() != 2:
        raise ValueError("Only tensors with 2 dimensions are supported")

    rows, cols = tensor.shape
    num_zeros = int(math.ceil(sparsity * rows))

    with torch.no_grad():
        tensor.normal_(0, std, generator=generator)
        for col_idx in range(cols):
            row_indices = torch.randperm(rows)
            zero_indices = row_indices[:num_zeros]
            tensor[zero_indices, col_idx] = 0
    return tensor


# for backward compatibility
def _make_deprecate(meth):
    new_name = meth.__name__
    old_name = new_name[:-1]

    def deprecated_init(*args, **kwargs):
        warnings.warn(
            f"`nn.init.{old_name}` is now deprecated in favor of `nn.init.{new_name}`.",
            FutureWarning,
            stacklevel=2,
        )
        return meth(*args, **kwargs)

    deprecated_init.__doc__ = rf"""
    {old_name}(...)

    .. warning::
        This method is now deprecated in favor of :func:`torch.nn.init.{new_name}`.

    See :func:`~torch.nn.init.{new_name}` for details."""
    deprecated_init.__name__ = old_name
    return deprecated_init


uniform = _make_deprecate(uniform_)
normal = _make_deprecate(normal_)
constant = _make_deprecate(constant_)
eye = _make_deprecate(eye_)
dirac = _make_deprecate(dirac_)
xavier_uniform = _make_deprecate(xavier_uniform_)
xavier_normal = _make_deprecate(xavier_normal_)
kaiming_uniform = _make_deprecate(kaiming_uniform_)
kaiming_normal = _make_deprecate(kaiming_normal_)
orthogonal = _make_deprecate(orthogonal_)
sparse = _make_deprecate(sparse_)