How to install pytorch 1.3.0 or above with cuda 8?

I have an old server with NVIDIA driver 375.26, which only supports cuda8.0 as indicated in CUDA Compatibility :: GPU Deployment and Management Documentation

The server has a P100 card, and I cannot upgrade the driver for various reasons.

How could I install PyTorch 1.3.0 or above with this setup?

The current min. CUDA version is 9.2 and CUDA8.0 is not tested, so I don’t know if you could be lucky in building it from source (have you tried it and was it failing?).

Thank you for your reply.

I haven’t tested building it from source. I decided to use the cpu version for now.