Function unique

Finds unique elements along an axis of a tensor.

It returns a tensor values containing all of the unique elements along the axis of the given tensor x in the same order that they occur along the axis in x; x does not need to be sorted. It also returns a tensor indices the same size as the number of the elements in x along the axis dimension. It contains the index in the unique output values.

// A 1-D tensor
const a = tf.tensor1d([1, 1, 2, 4, 4, 4, 7, 8, 8]);
const {values, indices} = tf.unique(a);
values.print();   // [1, 2, 4, 7, 8,]
indices.print();  // [0, 0, 1, 2, 2, 2, 3, 4, 4]
Copy

// A 2-D tensor with axis=0
//
// 'a' is: [[1, 0, 0],
//          [1, 0, 0],
//          [2, 0, 0]]
const a = tf.tensor2d([[1, 0, 0], [1, 0, 0], [2, 0, 0]]);
const {values, indices} = tf.unique(a, 0)
values.print();   // [[1, 0, 0],
                  //  [2, 0, 0]]
indices.print();  // [0, 0, 1]
Copy

// A 2-D tensor with axis=1
//
// 'a' is: [[1, 0, 0],
//          [1, 0, 0],
//          [2, 0, 0]]
const a = tf.tensor2d([[1, 0, 0], [1, 0, 0], [2, 0, 0]]);
const {values, indices} = tf.unique(a, 1)
values.print();   // [[1, 0],
                  //  [1, 0],
                  //  [2, 0]]
indices.print();  // [0, 1, 1]
Copy

Doc