【CN LLM】2.3Other Apps

【CN LLM】2.3Other Apps

AI  Learning Assistant No 1
AI Learning Assistant No 1
August 16th, 2023

  • Wenda:


Introduction: An LLM calling platform. Find and design automatic execution actions for the plug-in knowledge base of small models, and realize the generation ability no less than that of large models.

  • Jittor LLMs:

Address: https://www.hayo.com/entry/5451

Introduction: Large-scale model reasoning library for graph calculation: notebooks can run large models without a graphics card. It has the advantages of low cost, wide support, portability, and high speed.

  • LMFlow:


Introduction: LMFlow is a large model fine-tuning toolbox developed by the LMFlow team of Hong Kong University of Science and Technology. The LMFlow toolbox has the characteristics of strong scalability, high efficiency and convenience. LMFlow only uses 180K pieces of data fine-tuning to get the Robin model that ranks first on the Huggingface list. LMFlow supports users to quickly train personalized models, and it only takes a single 3090 and 5 hours to fine-tune a customized model with 7 billion parameters.

  • fastllm:

Address: https://www.hayo.com/entry/5453

Introduction: pure c++ full-platform llm acceleration library, chatglm-6B level model single card can reach 10000+token / s, support moss, chatglm, baichuan models, and run smoothly on the mobile phone.

  • WebCPM

Address: https://www.hayo.com/entry/5471

Introduction: A large Chinese model that supports interactive web search.

  • GPT Academic:

Address: https://www.hayo.com/entry/1805

Introduction: Provides a graphical interactive interface for GPT/GLM, especially optimizes the paper reading experience, supports parallel query of multiple LLM models, and supports local models such as Tsinghua chatglm. Compatible with Fudan MOSS, llama, rwkv, Pangu, etc.

  • ChatALL:

Address: https://www.hayo.com/entry/5478

Introduction: ChatALL (Chinese name: Qidao) can send a command to multiple AIs at the same time, which can help users find the best answer.

  • Creative Chat GLM:

Address: https://www.hayo.com/entry/5480

Introduction: You can use the functions of revision and continuation to generate creative content. You can use the "Continue" button to help ChatGLM think of a beginning and let it continue to generate more content. You can use the "Revision" button to modify the last sentence of ChatGLM reply.

  • docker-llama2-chat:

Address: https://www.hayo.com/entry/5481 

Introduction: Open source a rapid deployment solution that only takes three steps to get started with LLaMA2.

  • ChatGLM2-Voice-Cloning:

Address: https://www.hayo.com/entry/5482 

Introduction: Realized an application that can have an immersive conversation with a favorite character, mainly using the technology of ChatGLM2+voice cloning+video conversation.

Please pay attention to more content.

[CNLLM] Chinese large language model collection


no dataCoffee time! Feel free to comment