As diag, but returns a torch tensor (dense, to match
the CPU fallback's dense output) when an accelerator is active.
Usage
diag_mat(vec, backend = c("auto", "cpu", "gpu"))
Arguments
- vec
a numeric vector (or torch tensor).
- backend
one of "auto" (default), "cpu" or "gpu".
Value
a diagonal matrix (base matrix or torch tensor).
Examples
diag_mat(1:3, backend = "cpu")
#> [,1] [,2] [,3]
#> [1,] 1 0 0
#> [2,] 0 2 0
#> [3,] 0 0 3