FD75          ; mapped                 ; 0639 062C 0645 #1.1  ARABIC LIGATURE AIN WITH JEEM WITH MEEM FINAL FORM
FD76..FD77    ; mapped                 ; 0639 0645 0645 #1.1  ARABIC LIGATURE AIN WITH MEEM WITH MEEM FINAL FORM..ARABIC LIGATURE AIN WITH MEEM WITH MEEM INITIAL FORM
FD78          ; mapped                 ; 0639 0645 0649 #1.1  ARABIC LIGATURE AIN WITH MEEM WITH ALEF MAKSURA FINAL FORM

    of other integer types) in `tf.constant`.
*   Make the `gain` argument of convolutional orthogonal initializers
    (`convolutional_delta_orthogonal`, `convolutional_orthogonal_1D`,
    `convolutional_orthogonal_2D`, `convolutional_orthogonal_3D`) have
    consistent behavior with the `tf.initializers.orthogonal` initializer, i.e.
    scale the output l2-norm by `gain` and NOT by `sqrt(gain)`. (Note that these