Guides about rocm
- Install AMD ROCm on a GPU server ROCm is AMD's software stack for GPU computing. Register AMD's repository with amdgpu-install, install the kernel driver and ROCm, add your user to the GPU groups, reboot and verify. Intermediate 30 min
- Use your GPU in Docker containers: NVIDIA, AMD and Intel Containers do not see the GPU by default. NVIDIA GPUs need the NVIDIA Container Toolkit; AMD and Intel GPUs are passed in as devices. Set it up once and test it with a one-line container. Intermediate 20 min
- GPU troubleshooting: nvidia-smi errors, CUDA mismatches and memory Most GPU failures come down to the kernel module, the driver version or GPU memory. Read the exact message, match it below and apply the fix, from a reboot to a driver change. Intermediate 15 min