Skip to content

GPU Management¤

Note

Starting from v0.5, Lux has transitioned to a new GPU management system. The old system using cpu and gpu functions is still in place but will be removed in v0.6. Using the old functions might lead to performance regressions if used inside performance critical code.

Lux.jl can handle multiple GPU backends. Currently, the following backends are supported:

using Lux, LuxCUDA, LuxAMDGPU  # Important to load trigger packages

supported_gpu_backends()

("CUDA", "AMDGPU", "Metal")

Automatic Backend Management is done by two simple functions: cpu_device and gpu_device.

  1. cpu_device: This is a simple function and just returns a LuxCPUDevice object.
cdev = cpu_device()

(::LuxCPUDevice) (generic function with 2 methods)

x_cpu = randn(Float32, 3, 2)

3×2 Matrix{Float32}:
  0.433884   0.229779
 -0.459193  -1.95972
 -0.541064  -1.40102

  1. gpu_device: This function performs automatic GPU device selection and returns an object.

    1. If no GPU is available, it returns a LuxCPUDevice object.

    2. If a LocalPreferences file is present, then the backend specified in the file is used. To set a backend, use Lux.gpu_backend!(<backend_name>).

      1. If the trigger package corresponding to the device is not loaded, then a warning is displayed.
      2. If no LocalPreferences file is present, then the first working GPU with loaded trigger package is used.
gdev = gpu_device()

x_gpu = x_cpu |> gdev

3×2 CUDA.CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}:
  0.433884   0.229779
 -0.459193  -1.95972
 -0.541064  -1.40102

(x_gpu |> cdev) ≈ x_cpu

true

Manual Backend Management¤

Automatic Device Selection can be circumvented by directly using LuxCPUDevice and AbstractLuxGPUDevice objects.

cdev = LuxCPUDevice()

x_cpu = randn(Float32, 3, 2)

if LuxCUDA.functional()
    gdev = LuxCUDADevice()
    x_gpu = x_cpu |> gdev
elseif LuxAMDGPU.functional()
    gdev = LuxAMDGPUDevice()
    x_gpu = x_cpu |> gdev
else
    @info "No GPU is available. Using CPU."
    x_gpu = x_cpu
end

(x_gpu |> cdev) ≈ x_cpu

true