zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-11-04 10:38:10 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

docs fixes (#1263)

Browse Source
This commit is contained in:
Awni Hannun
2024-07-11 15:59:07 -07:00
committed by GitHub
parent d0da74209b
commit 218047c75a
3 changed files with 14 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
python/src/ops.cpp
View File

@@ -2559,7 +2559,7 @@ void init_ops(nb::module_& m) {
If unspecified, it defaults to ``-1``.
Returns:
array: The `uint32`` array containing indices that partition the input.
array: The ``uint32`` array containing indices that partition the input.
)pbdoc");
m.def(
"topk",
@@ -4388,20 +4388,21 @@ void init_ops(nb::module_& m) {
Perform the Walsh-Hadamard transform along the final axis.
Equivalent to:
```python
from scipy.linalg import hadamard
y = hadamard(len(x)) @ x
```
.. code-block:: python
Supports sizes `n = m*2^k` where m in (1, 12, 20, 28)
and 2^k <= 8192 for FP32 and 2^k <= 16384 for FP16/BF16.
from scipy.linalg import hadamard
y = hadamard(len(x)) @ x
Supports sizes ``n = m*2^k`` for ``m`` in ``(1, 12, 20, 28)`` and ``2^k
<= 8192`` for float32 and ``2^k <= 16384`` for float16/bfloat16.
Args:
a (array): Input array or scalar.
scale (float): Scale the output by this factor.
Returns:
array: The array with the new type.
array: The transformed array.
)pbdoc");
}

6
python/src/transforms.cpp
View File

@@ -934,9 +934,9 @@ void init_transforms(nb::module_& m) {
This class is meant to be used as a function decorator. Instances are
callables that behave identically to the wrapped function. However, when
a function transformation is used (e.g. computing gradients using
:func:`value_and_grad`) then the functions defined via :method:`vjp`,
:method:`jvp` and :method:`vmap` are used instead of the default
transformation.
:func:`value_and_grad`) then the functions defined via
:meth:`custom_function.vjp`, :meth:`custom_function.jvp` and
:meth:`custom_function.vmap` are used instead of the default transformation.
Note, all custom transformations are optional. Undefined transformations
fall back to the default behaviour.