Replication Data for: Temperature-Zero Does Not Guarantee Hardware-Invariant Determinism CPU vs GPU Inference Results for Llama 3 8B

Dataset containing prompt files, raw LLM outputs, and analysis scripts used to evaluate hardware-dependent divergence in temperature=0 greedy decoding for Llama 3 8B executed via Ollama.本数据集包含提示文件、原始大语言模型（LLM, Large Language Model）输出结果以及分析脚本，用于评估通过Ollama运行的Llama 3 8B模型在温度参数为0的贪心解码场景下的硬件相关解码差异。