Package index • Rmlx

Core API

as_mlx(): Create MLX array from R object
is_mlx(): Test if object is an MLX array
Rmlx-package Rmlx: Rmlx: R Interface to Apple's MLX Arrays
mlx-methods: Base R generics with mlx methods
mlx_dtype(): Get the data type of an MLX array
mlx_eval(): Force evaluation of an MLX operations
`[`(<mlx>) `[<-`(<mlx>): Subset MLX array
dim(<mlx>) mlx_shape(): Get dimensions of MLX array
`dim<-`(<mlx>): Set dimensions of MLX array
length(<mlx>): Get length of MLX array
print(<mlx>): Print MLX array
str(<mlx>): Object structure for MLX array
t(<mlx>): Transpose of MLX matrix
as.matrix(<mlx>): Convert MLX array to R matrix
as.array(<mlx>): Convert MLX array to R matrix/array
as.vector(<mlx>) as.logical(<mlx>) as.double(<mlx>) as.numeric(<mlx>) as.integer(<mlx>): Convert MLX array to R vector

Device and execution

mlx_default_device(): Get or set default MLX device
with_default_device(): Temporarily set the default MLX device
mlx_new_stream() mlx_default_stream(): MLX streams for asynchronous execution
mlx_set_default_stream(): Set the default MLX stream
mlx_best_device(): Get best available device
mlx_device(): Get device associated with an MLX object
mlx_has_gpu(): Check if GPU backend is available
mlx_synchronize(): Synchronize MLX execution
format(<mlx_stream>): Format method for mlx_stream
print(<mlx_stream>): Print method for mlx_stream

Creation

mlx_array(): Construct an MLX array from R data
mlx_matrix(): Construct MLX matrices efficiently
mlx_vector(): Construct MLX vectors
mlx_scalar(): Construct MLX scalars
mlx_zeros(): Create arrays of zeros on MLX devices
mlx_ones(): Create arrays of ones on MLX devices
mlx_zeros_like(): Zeros shaped like an existing mlx array
mlx_ones_like(): Ones shaped like an existing mlx array
mlx_full(): Fill an mlx array with a constant value
mlx_eye(): Identity-like matrices on MLX devices
mlx_identity(): Identity matrices on MLX devices
mlx_arange(): Numerical ranges on MLX devices
mlx_linspace(): Evenly spaced ranges on MLX devices
mlx_rand_bernoulli(): Sample Bernoulli random variables on mlx arrays
mlx_rand_categorical(): Sample from a categorical distribution on mlx arrays
mlx_rand_gumbel(): Sample from the Gumbel distribution on mlx arrays
mlx_rand_laplace(): Sample from the Laplace distribution on mlx arrays
mlx_rand_multivariate_normal(): Sample from a multivariate normal distribution on mlx arrays
mlx_rand_normal(): Sample from a normal distribution on mlx arrays
mlx_rand_permutation(): Generate random permutations on mlx arrays
mlx_rand_randint(): Sample random integers on mlx arrays
mlx_rand_truncated_normal(): Sample from a truncated normal distribution on mlx arrays
mlx_rand_uniform(): Sample from a uniform distribution on mlx arrays
mlx_key() mlx_key_split(): Construct MLX random number generator keys
mlx_key_bits(): Generate raw random bits on MLX arrays

Loading and saving

mlx_save(): Save an MLX array to disk
mlx_load(): Load an MLX array from disk
mlx_save_safetensors(): Save MLX arrays to the safetensors format
mlx_load_safetensors(): Load MLX arrays from the safetensors format
mlx_save_gguf(): Save MLX arrays to the GGUF format
mlx_load_gguf(): Load MLX tensors from the GGUF format
mlx_import_function(): Import an exported MLX function

Shape and indexing

mlx_reshape(): Reshape an mlx array
mlx_stack(): Stack mlx arrays along a new axis
mlx_squeeze(): Remove singleton dimensions
mlx_expand_dims(): Insert singleton dimensions
mlx_repeat(): Repeat array elements
mlx_tile(): Tile an array
mlx_pad() mlx_split(): Pad or split mlx arrays
mlx_roll(): Roll array elements
mlx_moveaxis() aperm(<mlx>): Reorder mlx array axes
mlx_contiguous(): Ensure contiguous memory layout
mlx_flatten(): Flatten axes of an mlx array
mlx_swapaxes(): Swap two axes of an mlx array
drop(): Drop singleton dimensions
row() col(): Row and column indices for mlx arrays
asplit(): Split mlx arrays along a margin
mlx_unflatten(): Unflatten an axis into multiple axes
mlx_meshgrid(): Construct coordinate arrays from input vectors
mlx_broadcast_to(): Broadcast an array to a new shape
mlx_broadcast_arrays(): Broadcast multiple arrays to a shared shape
mlx_where(): Elementwise conditional selection
mlx_tri() mlx_tril() mlx_triu(): Triangular helpers for MLX arrays
mlx_slice_update(): Update a slice of an mlx array
mlx_gather(): Gather elements from an mlx array
abind(): Bind mlx arrays along an axis
rbind(<mlx>): Row-bind mlx arrays
cbind(<mlx>): Column-bind mlx arrays

Ordering and selection

mlx_sort() mlx_argsort(): Sort and argsort for mlx arrays
mlx_topk() mlx_partition() mlx_argpartition(): Top-k selection and partitioning on mlx arrays
mlx_argmax() mlx_argmin(): Argmax and argmin on mlx arrays

Math and reductions

Math(<mlx>): Math operations for MLX arrays
Ops(<mlx>): Arithmetic and comparison operators for MLX arrays
Summary(<mlx>): Summary operations for MLX arrays
mlx_sum() mlx_prod() mlx_all() mlx_any() mlx_mean() mlx_var() mlx_std(): Reduce mlx arrays
mean(<mlx>): Mean of MLX array elements
mlx_cumsum() mlx_cumprod(): Cumulative sum and product
mlx_quantile() quantile(<mlx>): Compute quantiles of MLX arrays
mlx_clip(): Clip mlx array values into a range
mlx_maximum(): Elementwise maximum of two mlx arrays
mlx_minimum(): Elementwise minimum of two mlx arrays
mlx_hadamard_transform(): Hadamard transform for MLX arrays
mlx_softmax(): Softmax for mlx arrays
mlx_logsumexp(): Log-sum-exp reduction for mlx arrays
mlx_logcumsumexp(): Log cumulative sum exponential for mlx arrays
mlx_isnan() mlx_isinf() mlx_isfinite(): Elementwise NaN and infinity predicates
mlx_isposinf() mlx_isneginf(): Detect signed infinities in mlx arrays
mlx_nan_to_num(): Replace NaN and infinite values with finite numbers
mlx_real() mlx_imag() mlx_conjugate(): Complex-valued helpers for mlx arrays
mlx_degrees() mlx_radians(): Convert between radians and degrees
mlx_erf() mlx_erfinv(): Error function and inverse error function
mlx_isclose(): Element-wise approximate equality
mlx_allclose(): Test if all elements of two arrays are close
all(<equal.mlx>): Test if two MLX arrays are (nearly) equal
colSums(): Column sums for mlx arrays
rowSums(): Row sums for mlx arrays
colMeans(): Column means for mlx arrays
rowMeans(): Row means for mlx arrays
scale(<mlx>): Scale mlx arrays
fft(): Fast Fourier Transform
mlx_fft() mlx_fft2() mlx_fftn(): Fast Fourier transforms for MLX arrays

Probability distributions

mlx_dnorm() mlx_pnorm() mlx_qnorm(): Normal distribution functions
mlx_dunif() mlx_punif() mlx_qunif(): Uniform distribution functions
mlx_dexp() mlx_pexp() mlx_qexp(): Exponential distribution functions
mlx_dlnorm() mlx_plnorm() mlx_qlnorm(): Lognormal distribution functions
mlx_dlogis() mlx_plogis() mlx_qlogis(): Logistic distribution functions

Linear algebra

`%*%`(<mlx>): Matrix multiplication for MLX arrays
mlx_addmm(): Fused matrix multiply and add for MLX arrays
crossprod(<mlx>): Cross product
tcrossprod(<mlx>): Transposed cross product
outer(): Outer product of two vectors
diag(): Diagonal matrix extraction and construction
chol(<mlx>): Cholesky decomposition for mlx arrays
chol2inv(): Inverse from Cholesky decomposition
kronecker() kronecker.default() kronecker.mlx(): Kronecker product dispatcher
qr(<mlx>): QR decomposition for mlx arrays
svd(): Singular value decomposition
svd(<mlx>): Singular value decomposition for mlx arrays
solve(<mlx>): Solve a system of linear equations
pinv(): Moore-Penrose pseudoinverse for MLX arrays
mlx_kron(): Kronecker product for mlx arrays
mlx_inv(): Compute matrix inverse
mlx_tri_inv(): Compute triangular matrix inverse
mlx_cholesky_inv(): Compute matrix inverse via Cholesky decomposition
mlx_lu(): LU factorization
mlx_norm(): Matrix and vector norms for mlx arrays
mlx_solve_triangular() backsolve(): Solve triangular systems with mlx arrays
mlx_trace(): Matrix trace for mlx arrays
diag(<mlx>) mlx_diagonal(): Extract diagonal or construct diagonal matrix for mlx arrays
mlx_eig(): Eigen decomposition for mlx arrays
mlx_eigh(): Eigen decomposition of Hermitian mlx arrays
mlx_eigvals(): Eigenvalues of mlx arrays
mlx_eigvalsh(): Eigenvalues of Hermitian mlx arrays
mlx_cross(): Vector cross product with mlx arrays

Gradients and compilation

mlx_grad() mlx_value_grad(): Automatic differentiation for MLX functions
mlx_stop_gradient(): Stop gradient propagation through an mlx array
mlx_compile(): Compile an MLX Function for Optimized Execution
mlx_disable_compile() mlx_enable_compile(): Control Global Compilation Behavior

Neural networks

mlx_linear(): Create a learnable linear transformation
mlx_sequential(): Compose modules sequentially
mlx_set_training(): Toggle training mode for MLX modules
mlx_embedding(): Embedding layer
mlx_conv1d(): 1D Convolution
mlx_conv2d(): 2D Convolution
mlx_conv3d(): 3D Convolution
mlx_conv_transpose1d(): 1D Transposed Convolution
mlx_conv_transpose2d(): 2D Transposed Convolution
mlx_conv_transpose3d(): 3D Transposed Convolution
mlx_quantize(): Quantize a Matrix
mlx_dequantize(): Dequantize a Matrix
mlx_quantized_matmul(): Quantized Matrix Multiplication
mlx_gather_qmm(): Gather-based Quantized Matrix Multiplication
mlx_forward(): Forward pass utility

Activation functions

mlx_relu(): Rectified linear activation module
mlx_gelu(): GELU activation
mlx_sigmoid(): Sigmoid activation
mlx_tanh(): Tanh activation
mlx_silu(): SiLU (Swish) activation
mlx_leaky_relu(): Leaky ReLU activation
mlx_softmax_layer(): Softmax activation

Regularization and normalization

mlx_dropout(): Dropout layer
mlx_layer_norm(): Layer normalization
mlx_batch_norm(): Batch normalization

Loss functions

mlx_mse_loss(): Mean squared error loss
mlx_l1_loss(): L1 loss (Mean Absolute Error)
mlx_binary_cross_entropy(): Binary cross-entropy loss
mlx_cross_entropy(): Cross-entropy loss

Training utilities

mlx_parameters(): Collect parameters from modules
mlx_param_values(): Retrieve parameter arrays
mlx_param_set_values(): Assign arrays back to parameters
mlx_optimizer_sgd(): Stochastic gradient descent optimizer
mlx_train_step(): Single training step helper
mlx_coordinate_descent(): Coordinate Descent with L1 Regularization