Code source de src.module

import numpy as np
from typing import Any


class Loss(object):
    def forward(self, y, yhat):
        raise NotImplementedError()

    def backward(self, y, yhat):
        raise NotImplementedError()


class Module(object):
    def __init__(self):
        self._parameters = {}
        self._gradient = {}

    def __call__(self, *args: Any, **kwds: Any) -> Any:
        return self.forward(*args, **kwds)

    def calculate_gain(self):
        if self.__class__.__name__.lower() == "tanh":
            return 5 / 3
        elif self.__class__.__name__.lower() == "relu":
            return np.sqrt(2)
        else:
            return 1.0

    def zero_grad(self):
        """Réinitialise le gradient."""
        raise NotImplementedError()

    def forward(self, X):
        """Passe forward."""
        raise NotImplementedError()

    def update_parameters(self, learning_rate=1e-3):
        r"""Update the parameters according to the calculated gradient and the learning
        rate.
        """
        # self._parameters -= learning_rate * self._gradient
        raise NotImplementedError()

    def backward_update_gradient(self, input, delta):
        r"""Update gradient value given module.

        .. math::
            \frac{\partial L}{\partial w_i^h}=\sum_k \frac{\partial L}{\partial z_k^h} 
            \frac{\partial z_k^h}{\partial w_i^h}=\sum_k \delta_k^h 
            \frac{\partial z_k^h}{\partial w_i^h}, 
            \text { let } \nabla_{\mathbf{w}^h} L=\left(\begin{array}{ccc}
            \frac{\partial z_1^h}{\partial w_1^h} & \frac{\partial z_2^h}{\partial w_1^h} 
            & \cdots \\ \frac{\partial z_1^h}{\partial w_2^h} & \ddots & \\
            \vdots & \end{array}\right) \nabla_{\mathbf{z}^h L} 
        """
        raise NotImplementedError()

    def backward_delta(self, input, delta):
        r"""Calculates the derivative of the error and the next delta (derivative of the 
        module with respect to the to the inputs).

        .. math::
            \delta_j^{h-1}=\frac{\partial L}{\partial z_j^{h-1}}=\sum_k 
            \frac{\partial L}{\partial z_k^h} \frac{\partial z_k^h}{\partial z_j^{h-1}}, 
            \text { let } \nabla_{\mathbf{z}^{h-1}} L=\left(\begin{array}{ccc}
            \frac{\partial z_1^h}{z_1^{h-1}} & \frac{\partial z_2^h}{z_1^{h-1}} & \cdots \\
            \frac{\partial z_2^h}{z_2^{h-1}} & \ddots & \cdots \\
            \vdots & \end{array}\right) \nabla_{\mathbf{z}^h L}
        """
        raise NotImplementedError()