Ignore old GPU when building Pytorch?

[yahado]

I’ve a old which isn’t supported
(nvcc fatal : Unsupported gpu architecture ‘compute_30’),
and a new one which is fine.
Building CUDA it complains about the unsupported GPU, is there a way to ignore the old one?

[JosueCom (Josue)]

When choosing CUDA device, you can detrmine the GPU by doing torch.cuda.set_device([device]). Read more here; link.

Feel free to read this thread too for alternative options such as setting CUDA_VISIBLE_DEVICES

How to change the default device of GPU? device_ids[0]

How to change the default device of GPU? for some reason ,I can not use the device_ids[0] of GPU, I change the following code:（in data_parallel.py） if output_device is None: output_device =device_ids[0] to if output_device is None: output_device =device_ids[1] but it still seem to used the device_ids[0] all tensors must be on devices[0]？ How to change it？

[yahado]

Using CUDA_VISIBLE_DEVICES worked.