Chinese instruction tuning
According to the model card, this is a self-fine-tuned Chinese chat version of Llama-3-8B-Instruct built on Chinese NLP, dialogue, math, and ruozhiba Q&A sources.
Open Source Model Profile · Rookie
Llama-3-8B-Instruct-Chinese
Llama-3-8B-Instruct-Chinese is an 8.03B-parameter Llama Chinese chat fine-tune from Rookie. Its model card lists Chinese SFT datasets and documents a llama3 chat template.
- Publisher
- Rookie
- Task
- text-generation
- Model type
- llama
- License
- Unknown
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Llama-3-8B-Instruct-Chinese is published by Rookie as a Llama-family text-generation model. The captured configuration identifies LlamaForCausalLM, and Safetensors metadata reports 8,030,261,248 parameters (about 8.03B). According to the model card, it is a self-fine-tuned Chinese chat version of Llama-3-8B-Instruct trained on Chinese instruction, dialogue, math, and Q&A sources.
Recorded capabilities
Documented llama3 chat template
The card registers a llama3 template using Llama 3 system, user, and assistant header tokens with an eot stop word.
Quantization-tagged loading path
Hub tags include gguf, and the card documents 4-bit BitsAndBytes loading with optional PEFT adapter attachment.
Use cases in the source record
- Chinese conversational text generation using the card's documented llama3 chat template.
- Memory-efficient experiments that load the model in 4-bit BitsAndBytes mode with optional adapter weights.
Limitations and unknowns
- No license value was extracted for this record.
- No evaluation results were extracted from this record.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data, not independently refreshed current availability.
- Training claims rest on a code-heavy publisher card mixing Chinese prose with scripts and have not been independently verified by Ethen.
Source and provenance
Source: Rookie/Llama-3-8B-Instruct-Chinese
Captured: Unknown. Processed: 2026-09-07T19:34:36.288560+00:00.
Llama-3-8B-Instruct-Chinese-chat Llama-3-8B-Instruct in Chinese 自己微调版本 训练可用数据整理 数据集 介绍 firefly-train-1.1M 包含了23种常见的中文NLP任务的数据，并且构造了许多与中华文化相关的数据，如对联、作诗、文言文翻译、散文、金庸小说等。对于每个任务，由人工书写若干种指令模板，保证数据的高质量与丰富度，数据量为115万。 moss-003-sft-data 由复旦大学MOSS团队开源的中英文多轮对话数据，包含100万+数本。 school_math_0.25M 由BELLE项目组开源的数学运算指令数据，包含25万条数问。 ruozhiba 弱智吧数据问答，据说比较锻炼模型的心智能力。 欢迎补充，要求中文且一问一答形式，适合用于提升llama3任务能力的数据集 github地址 推荐微调工具 在此感谢以下项目，提供了许多优秀的中文微调工具，供大家参考： Firefly - https://github.com/yangjianxin1/Firefly LLaMA-Factory - https://github.com/hiyouga/LLaMA-Factory.git Chat版模型下载 Instruct + 继续中文sft版 huggingface地址 模型量化加速、部署 模型使用 默认情况下直接运行以下代码即可体验llama3中文对话，请自行修改 model_na…
F001F002F003F004F005F006F008F009F010F012F013