jax.numpy.unique_values#

jax.numpy.unique_values(x,/,*,size=None,fill_value=None)[source]#

Return unique values from x, along with indices, inverse indices, and counts.

JAX implementation ofnumpy.unique_values(); this is equivalent to callingjax.numpy.unique() withequal_nan set to True.

Because the size of the output ofunique_values is data-dependent, the functionis not typically compatible withjit() and other JAX transformations.The JAX version adds the optionalsize argument which must be specified staticallyforjnp.unique to be used in such contexts.

Parameters:

• x (ArrayLike) – N-dimensional array from which unique values will be extracted.

• size (int |None) – if specified, return only the firstsize sorted unique elements. If there are fewerunique elements thansize indicates, the return value will be padded withfill_value.

• fill_value (ArrayLike |None) – whensize is specified and there are fewer than the indicated number ofelements, fill the remaining entriesfill_value. Defaults to the minimum unique value.

Returns:

An arrayvalues of shape(n_unique,) containing the unique values fromx.

Return type:

Array

See also

• jax.numpy.unique(): general function for computing unique values.

• jax.numpy.unique_values(): compute onlyvalues.

• jax.numpy.unique_counts(): compute onlyvalues andcounts.

• jax.numpy.unique_inverse(): compute onlyvalues andinverse.

Examples

Here we compute the unique values in a 1D array:

>>>x=jnp.array([3,4,1,3,1])>>>jnp.unique_values(x)Array([1, 3, 4], dtype=int32)

For examples of thesize andfill_value arguments, seejax.numpy.unique().