conatus.expert.gpu_available

conatus.expert.gpu_available()[source]

True when backend='cuda' can run here: CuPy sees a device and the compiled core has CUDA with a device. gpu_report() says which part is missing.

Return type:

bool