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