AI Infrastructure Reference
Every AI Infrastructure Reference lesson in the Linux for AI Engineering course, in the order you study them. Each page opens about the first half of its lesson to read, some the whole of it; the full lessons are in the plan.
Appendix A: the command reference for AI infrastructure
The Linux commands for AI infrastructure that matter most, sorted by the question each one answers, layer by layer.
Appendix B: the troubleshooting playbook, from missing GPUs to NCCL hangs and Xid codes
A CUDA and GPU troubleshooting playbook: driver library mismatches after updates, checking layers from the bottom, Xid codes and hangs.
Appendix C: driver, CUDA, ROCm and framework compatibility matrices
Which CUDA, driver and PyTorch versions work together: the four things a wheel must match, checking every version, and the tables.
Appendix D: environment variables for CUDA, NCCL, PyTorch, Hugging Face, Ollama and vLLM
Why an exported variable is ignored by a service, how environment variables are inherited, where to set them, and CUDA and NCCL variables.
Appendix E: choosing hardware for AI workstations and servers
Choosing hardware for AI workstations and servers: GPU memory first, then memory speed, how GPUs connect, and the rest of the box.
Appendix F: the glossary of AI infrastructure terms
An AI infrastructure glossary: terms in plain words, and words like rank, kernel, node and cache that mean different things by field.
