Firefly-LLaMA2-Chinese is a bilingual Chinese-English LLM built on LLaMA-2 with Chinese vocabulary expansion. The project supports incremental pre-training of multiple model architectures (Baichuan2, Qwen, InternLM, Falcon, etc.) using QLoRA/LoRA techniques for low-resource training. It performs both continued pre-training and instruction fine-tuning on Chinese-English dialogue data, then evaluates on Open LLM Leaderboard and CMMLU benchmarks.