>>> import numpy as np
>>> import automatic_differentiation as ad
>>> x_val = np.random.randn(2, 3)
>>> w_val = np.random.randn(3, 5)
>>> x_val
array([[ 0.66588232, -1.24427652,  0.33879172],
       [-0.26769112,  0.52176526,  0.32342972]])
>>> w_val
array([[ 0.45380673,  0.21560083, -0.58899006,  0.59063819,  0.11651024],
       [-1.00214694,  2.86367365,  1.4583315 ,  0.08809848,  0.45746665],
       [ 0.20844987,  0.91073141,  0.07208344, -0.93793171, -0.72982744]])
>>> x = ad.Variable(x_val, name="x")
>>> w = ad.Variable(w_val, name="w")
>>> x
<automatic_differentiation.src.core.computational_graph.Variable object at 0x7fa30695f908>
>>> w
<automatic_differentiation.src.core.computational_graph.Variable object at 0x7fa30728a898>
>>> x()
array([[ 0.66588232, -1.24427652,  0.33879172],
       [-0.26769112,  0.52176526,  0.32342972]])
>>> w()
array([[ 0.45380673,  0.21560083, -0.58899006,  0.59063819,  0.11651024],
       [-1.00214694,  2.86367365,  1.4583315 ,  0.08809848,  0.45746665],
       [ 0.20844987,  0.91073141,  0.07208344, -0.93793171, -0.72982744]])
>>> y = x @ w
>>> y()
array([[ 1.61975087, -3.11108885, -2.18234443, -0.03408684, -0.7388924 ],
       [-0.5769466 ,  1.73100859,  0.94188804, -0.41549686, -0.02854644]])
>>> x_val @ w_val
array([[ 1.61975087, -3.11108885, -2.18234443, -0.03408684, -0.7388924 ],
       [-0.5769466 ,  1.73100859,  0.94188804, -0.41549686, -0.02854644]])
>>> w_grad = ad.grad(y, [w])[0]
>>> w_grad
<automatic_differentiation.src.core.ops.Add object at 0x7fa30689f0f0>
>>> w_grad()
array([[ 0.3981912 ,  0.3981912 ,  0.3981912 ,  0.3981912 ,  0.3981912 ],
       [-0.72251126, -0.72251126, -0.72251126, -0.72251126, -0.72251126],
       [ 0.66222144,  0.66222144,  0.66222144,  0.66222144,  0.66222144]])
>>> x_val.T @ np.ones_like(x_val @ w_val)
array([[ 0.3981912 ,  0.3981912 ,  0.3981912 ,  0.3981912 ,  0.3981912 ],
       [-0.72251126, -0.72251126, -0.72251126, -0.72251126, -0.72251126],
       [ 0.66222144,  0.66222144,  0.66222144,  0.66222144,  0.66222144]])
>>>