I want to have the real labels of SoftMax ‘output matrix and calculate the confusion matrix to know the performance of my model e.g. Accuracy for each label. The shape of y_pred is (1135,6) and y_test is (1135, ).

y_pred:
array([[5.63838069e-15, 4.49124792e-22, 1.70582300e-03, 9.98294175e-01,
    6.43675001e-16, 5.18157286e-08],
   [4.47809906e-22, 1.06405335e-19, 9.99843001e-01, 1.57030590e-04,
    2.76202466e-30, 4.41898809e-17],
   [1.71584970e-17, 1.95483637e-24, 3.00669862e-13, 1.00000000e+00,
    7.63759589e-19, 3.53933007e-18],
   ...,
   [1.90624938e-08, 0.00000000e+00, 1.78032694e-16, 2.00390571e-09,
    1.29498029e-15, 1.00000000e+00],
   [1.50917319e-19, 1.17262857e-16, 1.72155957e-14, 1.00000000e+00,
    0.00000000e+00, 5.56629790e-19],
   [0.00000000e+00, 4.34668734e-30, 5.76211607e-30, 1.00000000e+00,
    1.20244761e-33, 0.00000000e+00]], dtype=float32)
y_test:
array([2., 2., 3., ..., 5., 3., 3.])