29 — Hardware & GPU¶
Hardware and GPU computing: CUDA, acceleration, and memory optimization.
Scope¶
Cheat sheets for getting the most out of compute hardware. Framework training loops belong to 05-Deep-Learning.
Hardware-focused AI/ML cheat sheets for practitioners who need to understand GPU memory, CUDA, and LLM serving constraints.
Cheat Sheets¶
- GPU Basics — CPU vs GPU, VRAM, CUDA cores, bandwidth, parallelism
- NVIDIA CUDA — toolkit, runtime, driver model, streams, libraries
- CUDA Troubleshooting — availability, driver mismatch, OOM, slow GPU
- PyTorch GPU — device placement,
.to(), memory, autocast, multi-GPU - LLM Hardware — VRAM, KV cache, batching, quantization, fine-tuning
Status: 5 sheets complete.
Contribute¶
- Template: CHEATSHEET_TEMPLATE.md
- Guide: CONTRIBUTING.md