Large Model Training Lab is mainly based on large models and search enhancement generation, combined with large-scale corpus, to realize intelligent retrieval and generation of private domain knowledge, so as to improve the efficiency and accuracy of knowledge application. The training lab realizes multi-modal understanding based on pictures and texts uploaded by users, answers users' questions around uploaded materials, and then completes the collection and sorting of knowledge materials and the creation, training and application of knowledge base. The specific implementation contents include LLM, RAG, knowledge base system construction and knowledge base system training, which can support the ability training of artificial intelligence trainers, artificial intelligence application development engineers and other positions.
New Upgrade - Virtual Teaching Assistant("VTA")
On the basis of the original knowledge Q&A function based on LLM (large language model) and RAG (search enhancement generation), the VTA is generated by simulating video and audio of real people.
VTA:
- Non-real-time VTA
- Full-scene mimicry VTA (real-time VTA)
Applicable Majors: artificial intelligence/intelligent science and technology/computer related majors, while VTA is applicable to all majors.
Course Products: deep learning, natural language, AI large model foundation and application development, LangChain Foundation and application development
Project Products: installation, deployment and measurement of intelligent question answering system, Qwen3B and ChatGLM3-6B models based on big model Web
Application Scenarios: professional teaching, comprehensive practical training, competition training. VTA is applicable to IP creation, teaching video/short video production, digital live broadcast, cultural travel video, news media.
Feature
Keep up with hot topics
The project focuses on the design of large model and AIGC, and the practical application of LLM large model +RAG, and the real application scenario of enterprise based on knowledge base Agent.
Comprehensive technical coverage
It covers operating system, container, large model, RAG, application software, front-end UI development, docker principle and deployment, vector database, knowledge base application engineering and other technologies, improving students' skills in all aspects and strengthening the ability to combine professional courses with practice.
New technology easy to use
The core technology adopts the latest and mainstream LLM(large language model), RAG(search enhancement generation), Embedding, Vue, SpringBoot, and encapsulation friendly, low coupling, easy for students to get started.
Scene diversification
In addition to helping teachers carry out practical teaching, the system can also provide educational assistance, acting as a teaching assistant to answer students' common questions about courses and practical training around the clock. At the same time, online customer service can automatically answer customer inquiries to improve service efficiency and user satisfaction; In addition, it can also provide psychological counseling to understand and respond to the emotional needs of users.
VTA of low cost and high performance
VTA acts as personalized teaching assistant to make learning more lively and interesting. Compared with human assistants, the production cost of VTA is lower, and a model training can be used multiple times. The short video production capacity can be enhanced by over 10 times, facilitating both individual and collaborative usage. The production efficiency of VTA will continue to increase as more sound materials, text materials, video materials, and personal image materials are accumulated.
