Crate mlx_rs

Unofficial rust bindings for the MLX framework.

Modules

• builder
  Defines helper traits for builder pattern
• error
  Custom error types and handler for the c ffi
• fast
  Fast implementations of commonly used multi-op functions.
• fft
  Fast Fourier Transform (FFT) and its inverse (IFFT) for one, two, and N dimensions.
• linalg
  Linear algebra operations.
• losses
  Loss functions
• macros
  Macros for mlx-rs.
• module
  This mod defines the traits for neural network modules and parameters.
• nested
  Implements a nested hashmap
• nn
  Neural network support for MLX
• ops
  Operations
• optimizers
  Trait and implementations for optimizers.
• quantization
  Traits for quantization
• random
  Collection of functions related to random number generation
• transforms
  Implementations of function transformations.
• utils
  Utility functions and types.

Macros

• array
  A helper macro to create an array with up to 3 dimensions.
• assert_array_eq
  Asserts that two arrays are equal.

Structs

• Array
  An n-dimensional array.
• Device
  Representation of a Device in MLX.
• DtypeIter
  An iterator over the variants of Dtype
• Stream
  A stream of evaluation attached to a particular device.
• StreamOrDevice
  Parameter type for all MLX operations.

Enums

• DeviceType
  Type of device.
• Dtype
  Array element type

Traits

• ArrayElement
  A marker trait for array elements.
• FromNested
  A helper trait to construct Array from nested arrays or slices.
• FromScalar
  A helper trait to construct Array from scalar values.

Functions

• stop_gradient
  Stop gradients from being computed.
• stop_gradient_device
  Stop gradients from being computed.

Type Aliases

• complex64
  Type alias for num_complex::Complex<f32>.