Thousand Sails Large Model
Qianfan large-scale model platform is a one-stop large-scale model development and service operation platform for enterprise developers, and it is also the world's first one-stop enterprise-level large-scale model platform launched by Baidu Smart Cloud. On the basis of providing a full set of Wenxin large model services, it also supports third-party open source large models, various AI development tools and a complete development environment.
1. The most models in China
The open source large model is developing rapidly, and a large number of high-quality third-party models have emerged, but in different application scenarios, the efficiency is uneven. In order to allow enterprise users to quickly experience, test, and access services, the Qianfan large-scale model platform fully accesses 33 large-scale models such as the full series of LLaMA2, ChatGLM2, RWKV, MPT, Dolly, OpenLLaMA, and Falcon, and is the one with the most third-party large-scale models in China platform.
- Reliable models: The third-party large models hosted on the Qianfan large model platform are all high-quality models selected by the platform. Three major indicators are mainly evaluated: commercial availability, model effect, and model security.
- Stronger security: In order to ensure the security of the models used by enterprises and developers, Qianfan has made model security enhancements for all connected third-party models, which not only guarantees the content security of Wenxin large models, but also protects the third-party large models safety output.
- Lower cost: Qianfan has made secondary performance enhancements for each connected large model. By optimizing model throughput and reducing model size, the speed of model inference is greatly improved. According to estimates, after tuning, the model volume can be reduced to 25%-50%, and the inference cost can be reduced by 50%. This means that compared with direct calls, enterprises can greatly save costs and improve results by using these models on the Qianfan platform.
- Support a full set of tool chains and diverse model tuning: Qianfan has made in-depth adaptation to the large models that are connected, and provides a full set of tool chains for model retraining, and supports various forms of model tuning, including SFT (full parameter fine-tuning, Prompt Tuning, Lora) and reinforcement learning (reward model learning, reinforcement learning training), etc. Help enterprises and developers to quickly retrain based on the basic large model, and build an enterprise-specific large model.
Log in to the operation console of the Qianfan large model platform, and these large models can be deployed and called in the [Model Warehouse].
On this basis, developers can also choose a training mode suitable for their own task scenarios for tuning and training, so as to achieve the ideal model effect.
2. The most complete Prompt template in China
In order for the large model to quickly understand our needs, the Qianfan large model platform has released 103 Prompt templates, including more than ten scenarios such as dialogue, programming, e-commerce, medical care, games, translation, and speeches. Among them are not only the prompt templates based on Baidu Smart Cloud's accumulation in industrial practice, but also the prompt templates from Wenxin Yiyan's high-frequency users. These templates can be directly output to the large model, which greatly improves the interaction efficiency and output quality of the model.
Supplement: Prompt (prompt word) is to propose commands or instructions to the large model through natural language, instructing the large model to generate corresponding content. Prompt is very important for the use effect of large models, and high-quality prompt questions can get higher-quality answers. By optimizing and improving Prompt's questioning methods, the pertinence and accuracy of the model can be improved, thereby generating higher-quality content.
Log in to the operation console of the Qianfan large-scale model platform, and enter our question in the [Online Test] module: Is Ma Baoguo good?
Then introduce the Prompt template, plus our custom keywords
It will return a content that is more in line with expectations.
3. Summary
Qianfan large model platform covers the entire life cycle of large model development, including: data management, model training, evaluation & optimization, prediction service and prompt engineering, and helps users efficiently develop and deploy large model applications end-to-end, continuously reducing the size of the large model. The technical threshold allows more people and organizations to benefit from artificial intelligence.
Interested friends can click the link below to experience it first: