2023-12-05 12:10:03 -08:00
|
|
|
MLX
|
|
|
|
|
===
|
|
|
|
|
|
2023-12-05 14:18:20 -08:00
|
|
|
MLX is a NumPy-like array framework designed for efficient and flexible machine
|
|
|
|
|
learning on Apple silicon, brought to you by Apple machine learning research.
|
2023-12-05 12:10:03 -08:00
|
|
|
|
|
|
|
|
The Python API closely follows NumPy with a few exceptions. MLX also has a
|
|
|
|
|
fully featured C++ API which closely follows the Python API.
|
|
|
|
|
|
|
|
|
|
The main differences between MLX and NumPy are:
|
|
|
|
|
|
|
|
|
|
- **Composable function transformations**: MLX has composable function
|
|
|
|
|
transformations for automatic differentiation, automatic vectorization,
|
|
|
|
|
and computation graph optimization.
|
|
|
|
|
- **Lazy computation**: Computations in MLX are lazy. Arrays are only
|
|
|
|
|
materialized when needed.
|
|
|
|
|
- **Multi-device**: Operations can run on any of the supported devices (CPU,
|
|
|
|
|
GPU, ...)
|
|
|
|
|
|
2023-12-05 14:18:20 -08:00
|
|
|
The design of MLX is inspired by frameworks like `PyTorch
|
2023-12-05 12:10:03 -08:00
|
|
|
<https://pytorch.org/>`_, `Jax <https://github.com/google/jax>`_, and
|
2024-01-03 20:14:05 -08:00
|
|
|
`ArrayFire <https://arrayfire.org/>`_. A notable difference from these
|
2023-12-05 12:10:03 -08:00
|
|
|
frameworks and MLX is the *unified memory model*. Arrays in MLX live in shared
|
|
|
|
|
memory. Operations on MLX arrays can be performed on any of the supported
|
|
|
|
|
device types without performing data copies. Currently supported device types
|
|
|
|
|
are the CPU and GPU.
|
|
|
|
|
|
|
|
|
|
.. toctree::
|
|
|
|
|
:caption: Install
|
|
|
|
|
:maxdepth: 1
|
|
|
|
|
|
|
|
|
|
install
|
|
|
|
|
|
|
|
|
|
.. toctree::
|
|
|
|
|
:caption: Usage
|
|
|
|
|
:maxdepth: 1
|
|
|
|
|
|
2024-01-06 05:41:48 -08:00
|
|
|
usage/quick_start
|
2024-01-10 14:14:38 -08:00
|
|
|
usage/lazy_evaluation
|
2024-01-06 05:41:48 -08:00
|
|
|
usage/unified_memory
|
2024-01-10 14:14:38 -08:00
|
|
|
usage/indexing
|
|
|
|
|
usage/saving_and_loading
|
2024-01-17 17:15:29 -08:00
|
|
|
usage/function_transforms
|
2024-02-08 12:44:23 -08:00
|
|
|
usage/compile
|
2024-01-06 05:41:48 -08:00
|
|
|
usage/numpy
|
2024-06-06 20:28:06 -07:00
|
|
|
usage/distributed
|
2024-01-10 14:14:38 -08:00
|
|
|
usage/using_streams
|
2023-12-05 12:10:03 -08:00
|
|
|
|
|
|
|
|
.. toctree::
|
|
|
|
|
:caption: Examples
|
|
|
|
|
:maxdepth: 1
|
|
|
|
|
|
|
|
|
|
examples/linear_regression
|
|
|
|
|
examples/mlp
|
|
|
|
|
examples/llama-inference
|
|
|
|
|
|
|
|
|
|
.. toctree::
|
|
|
|
|
:caption: Python API Reference
|
|
|
|
|
:maxdepth: 1
|
|
|
|
|
|
|
|
|
|
python/array
|
2024-03-30 17:32:20 -07:00
|
|
|
python/data_types
|
2023-12-05 12:10:03 -08:00
|
|
|
python/devices_and_streams
|
|
|
|
|
python/ops
|
|
|
|
|
python/random
|
|
|
|
|
python/transforms
|
2024-03-30 17:32:20 -07:00
|
|
|
python/fast
|
2023-12-05 12:10:03 -08:00
|
|
|
python/fft
|
2024-01-03 20:14:05 -08:00
|
|
|
python/linalg
|
2024-03-14 12:46:45 -07:00
|
|
|
python/metal
|
2023-12-05 12:10:03 -08:00
|
|
|
python/nn
|
|
|
|
|
python/optimizers
|
2024-06-06 20:28:06 -07:00
|
|
|
python/distributed
|
2023-12-05 12:10:03 -08:00
|
|
|
python/tree_utils
|
|
|
|
|
|
|
|
|
|
.. toctree::
|
|
|
|
|
:caption: C++ API Reference
|
|
|
|
|
:maxdepth: 1
|
|
|
|
|
|
|
|
|
|
cpp/ops
|
|
|
|
|
|
|
|
|
|
.. toctree::
|
|
|
|
|
:caption: Further Reading
|
|
|
|
|
:maxdepth: 1
|
|
|
|
|
|
|
|
|
|
dev/extensions
|
2024-03-30 17:32:20 -07:00
|
|
|
dev/metal_debugger
|
2024-08-23 12:14:53 -07:00
|
|
|
dev/custom_metal_kernels
|
