Awesome AI Agents › Programming Languages

InternLM/xtuner

⭐ 5194 Python added to this list on 2025-04-19 repository created 2023-07-11

XTuner is an efficient, flexible, and full-featured toolkit designed for fine-tuning large language models (LLMs) and vision-language models (VLMs). It supports a wide range of models including InternLM2, Llama 2 and 3, Phi3, Qwen, Mistral, ChatGLM, Baichuan, and Mixtral, among others. XTuner is optimized for performance, capable of fine-tuning 7B parameter models on a single 8GB GPU and scaling up to multi-node training for models exceeding 70B parameters. It leverages high-performance operators like FlashAttention and Triton kernels to boost training throughput and integrates seamlessly with DeepSpeed for advanced ZeRO optimization techniques. The toolkit offers great flexibility, supporting various training algorithms such as QLoRA, LoRA, and full-parameter fine-tuning, allowing users to select the best approach for their needs. Its data pipeline is robust and adaptable, accommodating datasets in multiple formats, including open-source and custom datasets. XTuner also supports continuous pre-training, instruction fine-tuning, and agent fine-tuning, making it suitable for a wide range of machine learning workflows. XTuner facilitates chatting with large models using predefined templates and ensures that the fine-tuned models can be easily deployed using complementary tools like LMDeploy. It also supports large-scale evaluation frameworks such as OpenCompass and VLMEvalKit. The project is actively maintained with frequent updates adding support for new models, training methods, and performance improvements. It includes benchmarks demonstrating training speed on various model sizes and provides extensive documentation and community support through platforms like GitHub, Discord, Twitter, and WeChat. Overall, XTuner is a comprehensive solution for researchers and developers aiming to fine-tune and deploy state-of-the-art LLMs and VLMs efficiently and effectively, with a strong emphasis on scalability, flexibility, and ease of use.

https://github.com/InternLM/xtuner

agentagent-fine-tuningbaichuanchatbotchatglmchatglm2chatglm3continuous-pre-trainingconversational-aidata-pipelinedeepspeedefficientfine-tuningflashattentionflexiblefull-parameter-fine-tuninginstruction-fine-tuninginternlminternlm2large-language-modelsllama-2llama-3llama2llama3llavallmllm-traininglmdeployloramistralmixtralmsagentmulti-node-trainingopencompasspeftphi3qloraqwenscalablesupervised-finetuningtraining-speed-benchmarktriton-kernelsvision-language-modelsvlmvlmevalkit