分享如何训练、评估大型语言模型,基于RAG、Agent、Chain构建有趣的LLMs应用。
- Mistral-8x7b-Instruct 稳定输出Json Format, 搭配Llamacpp grammar
- Mistral-8x7b-Instruct CoT Agent, Think step by steps
- Mistral-8x7b-Instruct ReAct Agent with tool call
- Llama3-8b-Instruct, transformers, vLLM and Llamacpp多种方法调戏
- Llama3-8b-Instruct, CoT with vLLM
- Llama3-8b-Instruct, 纯中文实现ReAct with tool call
- Chinese-Llama3-8b, DPO微调让Llama3更愿意说中文
- llama-cpp-convert-GGUF, 模型量化转化为GGUF格式并上传huggingface
Mixtral 8x7b ReAct | Llama3-8b ReAct |
---|---|
![]() |
![]() |