How to correctly configure the GPU in the qlib configuration file

[kitcoun]

❓ Questions and Help

rdagent been talking about GPU-related topics in the chat conversation.

My qlib configuration settings:
`
task:
.....

 GPU: 0 

`

(base) ubuntu@PMS:~$ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Tue_May_27_02:21:03_PDT_2025
Cuda compilation tools, release 12.9, V12.9.86
Build cuda_12.9.r12.9/compiler.36037853_0

The system has long had CUDA installed. Do I not need to install CUDA myself? Will the code install it automatically? Or do I need to configure something separately?

[notes4mark]

You should install CUDA version Torch by yourself.

[SunsetWolf]

Hi, @kitcoun
Thank you for your attention to RD-Agent. Neither RD-Agent nor Qlib will automatically install CUDA. If your system already has CUDA installed, you don’t need to install it again. However, you do need to make sure that:

1. Your PyTorch is the GPU version, not CPU-only;
2. PyTorch's CUDA version is compatible with your system's CUDA driver;
3. The GPU: 0 in Qlib's configuration just tells it to use the 0th GPU, you need to make sure that GPU is visible on the system.