Function xavier_normal_

python/oneflow/nn/init.py:117–145

r""" Fills the input `Tensor` with values according to the method described in `Understanding the difficulty of training deep feedforward neural networks` - Glorot, X. & Bengio, Y. (2010), using a normal distribution. The resulting tensor will have values sampled from :math:`\mat

(tensor, gain=1.0, *, data_format="NCHW")

def xavier_normal_(tensor, gain=1.0, *, data_format="NCHW"):
    r"""
    Fills the input `Tensor` with values according to the method
    described in `Understanding the difficulty of training deep feedforward
    neural networks` - Glorot, X. & Bengio, Y. (2010), using a normal
    distribution. 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}}}

    The interface is consistent with PyTorch.
    The documentation is referenced from: https://pytorch.org/docs/1.10/nn.init.html.

    Also known as Glorot initialization.

    Args:
        tensor: an n-dimensional `oneflow.Tensor`
        gain: an optional scaling factor

    Examples:
        >>> w = flow.empty(3, 5)
        >>> nn.init.xavier_normal_(w)
    """
    if os.getenv("ONEFLOW_ENABLE_NHWC") == "1":
        data_format = "NHWC"
    fan = calc_fan(tensor.shape, "fan_sum", get_data_format(data_format))
    std = gain * math.sqrt(2.0 / fan)
    return normal_(tensor, 0.0, std)


def orthogonal_(tensor, gain=1.0):