I am currently working on an HPC which allows me to install only cuda 12.2 and cuda 12.4 via spack system OS Red Hat Enterprise Linux 8.7 (GCC 12.2, glibc 2.28). I was wondering if there is a way to build from source Pytorch 2.7.1 with CUDA12.4. Unfortunately I can’t use conda so I have to rely on pip or building from source.
Thanks in advance for the help.
I’m unsure if I understand your issue correctly, but please note that your locally installed CUDA toolkit won’t be used if you install our PyTorch binaries as they ship with their own CUDA runtime dependencies. You need to properly install an NVIDIA driver and can then install any of our PyTorch binaries which support your GPUs.
I am currently using a CUDA12.4 compiler installed via spack, I’m working on a node with 4 nvidia A100-SXM-64GB,
driver info:
NVIDIA-SMI 535.54.03 Driver Version: 535.54.03 CUDA Version: 12.2