docs update

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

@@ -1,5 +1,5 @@
-Developer Documentation
-=======================
+Custom Extensions in MLX
+========================
 
 You can extend MLX with custom operations on the CPU or GPU. This guide
 explains how to do that with a simple example.
@@ -494,7 +494,7 @@ below.
         auto kernel = d.get_kernel(kname.str(), "mlx_ext");
 
         // Prepare to encode kernel
-        auto compute_encoder = d.get_command_encoder(s.index);
+        auto& compute_encoder = d.get_command_encoder(s.index);
         compute_encoder->setComputePipelineState(kernel);
 
         // Kernel parameters are registered with buffer indices corresponding to
@@ -503,11 +503,11 @@ below.
         size_t nelem = out.size();
 
         // Encode input arrays to kernel
-        set_array_buffer(compute_encoder, x, 0);
-        set_array_buffer(compute_encoder, y, 1);
+        compute_encoder.set_input_array(x, 0);
+        compute_encoder.set_input_array(y, 1);
 
         // Encode output arrays to kernel
-        set_array_buffer(compute_encoder, out, 2);
+        compute_encoder.set_output_array(out, 2);
 
         // Encode alpha and beta
         compute_encoder->setBytes(&alpha_, sizeof(float), 3);
@@ -531,7 +531,7 @@ below.
 
         // Launch the grid with the given number of threads divided among
         // the given threadgroups
-        compute_encoder->dispatchThreads(grid_dims, group_dims);
+        compute_encoder.dispatchThreads(grid_dims, group_dims);
     }
 
 We can now call the :meth:`axpby` operation on both the CPU and the GPU!
@@ -825,7 +825,7 @@ Let's look at a simple script and its results:
 
     print(f"c shape: {c.shape}")
     print(f"c dtype: {c.dtype}")
-    print(f"c correctness: {mx.all(c == 6.0).item()}")
+    print(f"c correct: {mx.all(c == 6.0).item()}")
 
 Output:
 

docs/build/html/_sources/install.rst

@@ -153,6 +153,8 @@ should point to the path to the built metal library.
      - OFF
    * - MLX_BUILD_METAL
      - ON
+   * - MLX_BUILD_CPU
+     - ON
    * - MLX_BUILD_PYTHON_BINDINGS
      - OFF
    * - MLX_METAL_DEBUG
@@ -179,10 +181,28 @@ should point to the path to the built metal library.
 
       xcrun -sdk macosx --show-sdk-version
 
+Binary Size Minimization
+~~~~~~~~~~~~~~~~~~~~~~~~
+
+To produce a smaller binary use the CMake flags `CMAKE_BUILD_TYPE=MinSizeRel`
+and `BUILD_SHARED_LIBS=ON`.
+
+The MLX CMake build has several additional options to make smaller binaries.
+For example, if you don't need the CPU backend or support for safetensors and
+GGUF, you can do:
+
+.. code-block:: shell
+
+  cmake ..
+    -DCMAKE_BUILD_TYPE=MinSizeRel \
+    -DBUILD_SHARED_LIBS=ON \
+    -DMLX_BUILD_CPU=ON \
+    -DMLX_BUILD_SAFETENSORS=OFF \
+    -DMLX_BUILD_GGUF=OFF
+
 Troubleshooting
 ^^^^^^^^^^^^^^^
 
-
 Metal not found
 ~~~~~~~~~~~~~~~
 

docs/build/html/_sources/python/_autosummary/mlx.core.addmm.rst

@@ -0,0 +1,6 @@
+mlx.core.addmm
+==============
+
+.. currentmodule:: mlx.core
+
+.. autofunction:: addmm

docs/build/html/_sources/python/_autosummary/mlx.core.as_strided.rst

@@ -0,0 +1,6 @@
+mlx.core.as\_strided
+====================
+
+.. currentmodule:: mlx.core
+
+.. autofunction:: as_strided

docs/build/html/_sources/python/_autosummary/mlx.core.linalg.cholesky.rst

@@ -0,0 +1,6 @@
+mlx.core.linalg.cholesky
+========================
+
+.. currentmodule:: mlx.core.linalg
+
+.. autofunction:: cholesky

docs/build/html/_sources/python/_autosummary/mlx.core.linalg.inv.rst

@@ -0,0 +1,6 @@
+mlx.core.linalg.inv
+===================
+
+.. currentmodule:: mlx.core.linalg
+
+.. autofunction:: inv

docs/build/html/_sources/python/_autosummary/mlx.core.linalg.svd.rst

@@ -0,0 +1,6 @@
+mlx.core.linalg.svd
+===================
+
+.. currentmodule:: mlx.core.linalg
+
+.. autofunction:: svd

docs/build/html/_sources/python/_autosummary/mlx.core.power.rst

@@ -0,0 +1,6 @@
+mlx.core.power
+==============
+
+.. currentmodule:: mlx.core
+
+.. autofunction:: power

docs/build/html/_sources/python/_autosummary/mlx.core.remainder.rst

@@ -0,0 +1,6 @@
+mlx.core.remainder
+==================
+
+.. currentmodule:: mlx.core
+
+.. autofunction:: remainder

docs/build/html/_sources/python/linalg.rst

@@ -8,5 +8,8 @@ Linear Algebra
 .. autosummary:: 
    :toctree: _autosummary 
 
+    inv
     norm
+    cholesky
     qr
+    svd

docs/build/html/_sources/python/nn/_autosummary/mlx.nn.Conv3d.rst

@@ -0,0 +1,16 @@
+mlx.nn.Conv3d
+=============
+
+.. currentmodule:: mlx.nn
+
+.. autoclass:: Conv3d
+
+   
+
+   
+   .. rubric:: Methods
+
+   .. autosummary::
+   
+   
+   

docs/build/html/_sources/python/nn/_autosummary/mlx.nn.Embedding.rst

@@ -13,5 +13,6 @@
    .. autosummary::
    
          ~Embedding.as_linear
+         ~Embedding.to_quantized
    
    

docs/build/html/_sources/python/nn/_autosummary/mlx.nn.Linear.rst

@@ -12,5 +12,6 @@
 
    .. autosummary::
    
+         ~Linear.to_quantized
    
    

docs/build/html/_sources/python/nn/layers.rst

@@ -15,6 +15,7 @@ Layers
    BatchNorm
    Conv1d
    Conv2d
+   Conv3d
    Dropout
    Dropout2d
    Dropout3d

docs/build/html/_sources/python/ops.rst

@@ -10,6 +10,7 @@ Operations
 
    abs
    add
+   addmm
    all
    allclose
    any
@@ -26,6 +27,7 @@ Operations
    argpartition
    argsort
    array_equal
+   as_strided
    atleast_1d
    atleast_2d
    atleast_3d
@@ -76,6 +78,7 @@ Operations
    isnan
    isneginf
    isposinf
+   issubdtype
    left_shift
    less
    less_equal
@@ -106,11 +109,13 @@ Operations
    outer
    partition
    pad
+   power
    prod
    quantize
    quantized_matmul
    radians
    reciprocal
+   remainder
    repeat
    reshape
    right_shift