Model Introduction

Compared with the previous generation CogVLM open source model, the CogVLM2 series open source model has the following improvements:

• Significant improvements on many benchmarks, such as TextVQA, DocVQA.
• Supports 8K content length.
• Supports image resolution up to 1344*1344.
• An open source model version is provided, supporting Chinese and English.

How to run

1. Open a new terminal page
2. Enter the command

cd CogVLM2/basic_demo
chainlit run web_demo.py --port 8080

Run the program

3. Copy the API address on the right and open it in your browser 
4. Click below to upload a picture and enter the text prompt

5. Press Enter to generate the answer