docs

docs/build/html/_sources/dev/extensions.rst

@@ -131,7 +131,7 @@ back and go to our example to give ourselves a more concrete image.
         * A primitive must know how to evaluate itself on the CPU/GPU
         * for the given inputs and populate the output array.
         *
-        * To avoid unecessary allocations, the evaluation function
+        * To avoid unnecessary allocations, the evaluation function
         * is responsible for allocating space for the array.
         */
         void eval_cpu(const std::vector<array>& inputs, array& out) override;
@@ -945,4 +945,4 @@ Scripts
 .. _Metal-cpp: https://developer.apple.com/metal/cpp/
 .. _`Metal Specification`: https://developer.apple.com/metal/Metal-Shading-Language-Specification.pdf
 .. _`Metal Example`: https://developer.apple.com/documentation/metal/performing_calculations_on_a_gpu?language=objc
-.. _PyBind11: https://pybind11.readthedocs.io/en/stable/
+.. _PyBind11: https://pybind11.readthedocs.io/en/stable/

docs/build/html/_sources/examples/llama-inference.rst

@@ -321,7 +321,7 @@ which can then be used to update the model. Note that the method above incurs
 several unnecessary copies from disk to numpy and then from numpy to MLX. It
 will be replaced in the future with direct loading to MLX.
 
-You can download the full example code in `mlx-examples <code>`_. Assuming, the
+You can download the full example code in `mlx-examples`_. Assuming, the
 existence of ``weights.pth`` and ``tokenizer.model`` in the current working
 directory we can play around with our inference script as follows (the timings
 are representative of an M1 Ultra and the 7B parameter Llama model):
@@ -369,9 +369,9 @@ Scripts
 
 .. admonition:: Download the code
 
-   The full example code is available in `mlx-examples <code>`_.
+   The full example code is available in `mlx-examples`_.
 
-.. code: `https://github.com/ml-explore/mlx-examples/tree/main/llama`_
+.. _mlx-examples: https://github.com/ml-explore/mlx-examples/tree/main/llama
 
 .. [1] Su, J., Lu, Y., Pan, S., Murtadha, A., Wen, B. and Liu, Y., 2021.
    Roformer: Enhanced transformer with rotary position embedding. arXiv

docs/build/html/_sources/examples/mlp.rst

@@ -127,5 +127,5 @@ Finally, we put it all together by instantiating the model, the
   This should not be confused with :func:`mlx.core.value_and_grad`.
 
 The model should train to a decent accuracy (about 95%) after just a few passes
-over the training set. The `full example <https://github.com/ml-explore/mlx-examples/tree/main/mlp>`_
+over the training set. The `full example <https://github.com/ml-explore/mlx-examples/tree/main/mnist>`_
 is available in the MLX GitHub repo.

docs/build/html/_sources/install.rst

@@ -11,6 +11,10 @@ silicon computer is
 
     pip install mlx
 
+.. note::
+    MLX is only available on devices running MacOS >= 13.3 
+    It is highly recommended to use MacOS 14 (Sonoma)
+
 Build from source
 -----------------
 
@@ -19,6 +23,7 @@ Build Requirements
 
 - A C++ compiler with C++17 support (e.g. Clang >= 5.0)
 - `cmake <https://cmake.org/>`_ -- version 3.24 or later, and ``make``
+- Xcode >= 14.3 (Xcode >= 15.0 for MacOS 14 and above)
 
 
 Python API
@@ -111,3 +116,21 @@ should point to the path to the built metal library.
      - ON
    * - MLX_BUILD_PYTHON_BINDINGS
      - OFF
+
+
+.. note::
+
+    If you have multiple Xcode installations and wish to use 
+    a specific one while building, you can do so by adding the 
+    following environment variable before building 
+
+    .. code-block:: shell
+
+      export DEVELOPER_DIR="/path/to/Xcode.app/Contents/Developer/"
+
+    Further, you can use the following command to find out which 
+    MacOS SDK will be used
+
+    .. code-block:: shell
+
+      xcrun -sdk macosx --show-sdk-version