0.1 Understanding the high-level LLM training pipeline (pretraining → finetuning → alignment) 0.2 Hardware & software environment setup (PyTorch, CUDA/Mac, mixed precision, profiling tools) conda ...