sagemaker-llm-training-skill

Standard Operating Procedure (SOP) for training and fine-tuning LLMs on Amazon SageMaker. Use when the user wants to: fine-tune, continued pretraining, CPT, preference optimization, DPO, RLHF, LoRA, QLoRA, Spectrum, full fine-tuning, SFT on SageMaker. Also use when choosing between SageMaker Training Jobs vs HyperPod clusters, selecting training containers (DLC), or GPU vs Trainium instances. Generates runnable notebooks or Python scripts. Triggers: "train llm", "fine-tune", "sagemaker training", "lora training", "qlora", "hyperpod training", "trainium training", "dlc container", "training job".