A **tensor** is a data structure that represents matrices and vectors. It can be thought of as a multi-dimensional array. 

The **`unique` function** in TensorFlow is used to find unique values in a tensor.
It takes the original tensor `x`. Then, it returns two tensors: 
1. `y` 
2.  `idx`

Tensor `y` contains the unique values from the original tensor. Tensor `idx` contains the index position where each unique value is found.

> Tensor `y` has the same order of elements as in Tensor `x`. 

> Tensor `idx` has the same number of elements as in tensor `x`.

The illustration below shows how the `unique` function works in TensorFlow:

# Syntax

The syntax of the `unique` function is as follows:

```python
tf.unique(x, out_idx=tf.dtypes.int32, name=None)
```

# Parameters

The `unique` function has three parameters:

* `x`: A one-dimensional tensor.
* `out_idx`: Data type of the return tensor. Possible values include `tf.int32` and `tf.int64`. It is `tf.int32` by default.
* `name`: A name for the operation.

> Only the first parameter is compulsory. The rest are optional.

# Return values

The `unique` function returns two tensors: `y` and `idx`. 

Tensor `y` contains the unique values from the original tensor. Tensor `idx` contains the index position where each unique value is found.

# Example

The code snippet below shows how the `unique` function is used in TensorFlow:

import tensorflow as tf

# tensor 'x' is [1, 1, 2, 4, 4, 4, 7, 8, 8]
x = tf.constant([1,1,2,4,4,4,7,8,8])
y, idx = tf.unique(x)
print("x:", x) # => [1, 1, 2, 4, 4, 4, 7, 8, 8]
print("y:", y) # => [1, 2, 4, 7, 8]
print("idx", idx) # => [0, 0, 1, 2, 2, 2, 3, 4, 4]

What is the unique function in TensorFlow?

Finds unique values in a tensor, returning unique values and their index positions.