Skip to contents

Enable or disable GPU acceleration

Source: R/torch-setup.R
use_gpu.Rd

Sets package-wide options controlling whether optimize_alpha() uses GPU-accelerated optimization (PB-SGD) or CPU-only methods.

Usage

use_gpu(enable = TRUE, device = NULL, dtype = "float32")

Arguments

enable

Logical. TRUE to enable GPU, FALSE for CPU only.

device

Character or NULL. Torch device: "cuda", "mps", or "cpu". NULL = auto-detect.

dtype

Character. "float32" or "float64". Default: "float32".

Value

Invisibly returns the previous settings as a list.

Details

When GPU is enabled, optimize_alpha() uses torch-based PB-SGD (per-bracket stochastic gradient descent). This requires the torch package to be installed with a working GPU backend (CUDA for NVIDIA, MPS for Apple Silicon).

The per-call use_gpu parameter in optimize_alpha() always overrides the global setting.

See also

check_torch() to diagnose GPU setup, setup_torch() to install torch with GPU support.

Examples

use_gpu(FALSE) # ensure CPU mode
#> GPU disabled. Using CPU optimization.