Artificial intelligence (AI) is rapidly evolving, and language models (LMs) are becoming increasingly capable of helping us solve complex AI tasks. As the complexity of AI tasks increases, so does the need for LMs to interface with numerous AI models. This is where HuggingGPT comes in. In this article, we'll take a closer look at HuggingGPT and how it can help you solve complex AI tasks.
 |
| HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in HuggingFace |
HuggingGPT is a collaborative system that consists of an LLM as the controller and numerous expert models as collaborative executors. The workflow of the HuggingGPT system consists of four stages: Task Planning, Model Selection, Task Execution, and Response Generation. Let's take a closer look at each of these stages.
Task Planning
The first stage of the HuggingGPT system is Task Planning. Using ChatGPT, HuggingGPT analyzes the requests of users to understand their intention, and disassemble them into possible solvable tasks. This allows the system to better understand what the user is looking for and to plan accordingly.
Model Selection
Once the task has been planned, HuggingGPT moves on to the Model Selection stage. To solve the planned tasks, ChatGPT selects expert models hosted on Hugging Face based on their descriptions. This ensures that the system is using the best models available for the task at hand.
Task Execution
With the models selected, HuggingGPT moves on to the Task Execution stage. In this stage, the system invokes and executes each selected model, and returns the results to ChatGPT. This ensures that the system is using the best models available for the task at hand.
Response Generation
Finally, using ChatGPT to integrate the prediction of all models, HuggingGPT moves on to the Response Generation stage. In this stage, the system generates responses that take into account the predictions made by each model. This ensures that the system is providing the user with the best possible response to their request.
 |
| HuggingGPT inputs |
 |
| HuggingGPT Response |
System Requirements
To use HuggingGPT, you'll need to make sure your system meets the minimum requirements. The default requirements for HuggingGPT are:
Ubuntu 16.04 LTS
VRAM >= 12GB
RAM > 12GB (minimal), 16GB (standard), 42GB (full)
Disk > 78G (with 42G for damo-vilab/text-to-video-ms-1.7b)
If you don't meet these requirements, don't worry. The configuration lite.yaml does not require any expert models to be downloaded and deployed locally. However, it means that Jarvis is restricted to models running stably on HuggingFace Inference Endpoints.
Quick Start
To get started with HuggingGPT, you'll need to replace openai.key and huggingface.token in server/config.yaml with your personal OpenAI Key and your Hugging Face Token.
To read more, check their official page.