Base R generics with mlx methods

Rmlx provides S3 methods for a number of base R generics so that common operations keep working after converting objects with as_mlx(). The main entry points are:

Details

• %*% for matrix multiplication

• [ and [<- for extraction and assignment

• Ops and Math for elementwise arithmetic and math

• Summary for reductions such as sum() and max(); also mean(), length() and all.equal().

• diag(), dim() and dim<-

• as.matrix(), as.array(), and as.vector() for conversion back to base R

• row() and col() for index helpers that play nicely with mlx arrays

• cbind() and rbind() for binding arrays along rows or columns; there is also an abind() function modelled on abind::abind().

• rowMeans(), colMeans(), rowSums(), and colSums() for axis-wise summaries

• aperm(), t(), and dim<- for shape manipulation

• kronecker(), outer(), crossprod(), and tcrossprod() for linear algebra helpers

• fft(), chol(), chol2inv(), backsolve(), and solve() for numerical routines

• scale() for column-wise centring and scaling that stays on the MLX backend

• asplit() to slice arrays along a margin while staying on the MLX backend

• is.finite(), is.infinite() and is.nan()

Most methods return mlx objects. One exception is that all() and any() return standard R TRUE or FALSE when used on mlx objects.

See also

as_mlx()