SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI developer platformsChatGLM-webui

ChatGLM-webui

ChatGLM-webui is an open-source project that provides a web interface for the ChatGLM-6B model. It was developed by Akegarasu and uses Gradio Chatbox to build the chat interface, offering a better user experience than the original demo. The project provides a one-click installation script (Python still needs to be installed separately), supports adjusting more parameters, conveniently saving and loading conversation history and presets, and allows customizing the maximum context length. It also supports saving conversations as Markdown files. At startup, command-line arguments can be used to specify the model path, listening address, port, sharing method, computation precision, and whether to use CPU, etc.

Open website

About this tool

ChatGLM-webui is a web interface (WebUI) project for the ChatGLM-6B model developed by THUDM, published on GitHub. Its goal is to use Gradio Chatbox for presentation instead of the demo-like interface of the original ChatGLM-6B, thereby providing a better user experience.

Main capabilities and features:

  • Uses Gradio Chatbox to build the chat interface, with a better interactive experience than the original demo.
  • Provides a one-click installation script, but users still need to install Python in advance.
  • Provides more freely adjustable parameters than the original version.
  • Supports conveniently saving and loading conversation history and presets.
  • Supports customizing the maximum context length.
  • Supports saving conversation content as Markdown files.

Usage (via program arguments):

  • --model-path: Specify the model path. If not manually specified, the default value is THUDM/chatglm-6b, and Transformers will automatically download the model from Hugging Face.
  • --listen: Start Gradio with 0.0.0.0 as the server name, allowing it to respond to network requests.
  • --port: Specify the WebUI port.
  • --share: Use Gradio's sharing feature.
  • --precision: Options are fp32 (CPU only), fp16, int4 (CUDA GPU only), int8 (CUDA GPU only).
  • --cpu: Run using CPU.
  • --path-prefix: URL root path. If not manually specified, the default value is /. After using /foo/bar as the path prefix, ChatGLM-webui will be served from http://$ip:$port/foo/bar/ instead of http://$ip:$port/。

Applicable scenarios: This project is suitable for users who want to set up an interactive web chat interface for the ChatGLM-6B model locally or on a server, providing conveniences such as conversation history management, parameter adjustment, and Markdown export.

Limitations: The one-click installation script still requires Python to be installed in advance; int4 and int8 precision only support CUDA GPUs, and fp32 only supports CPU. Other functions, pricing details, and specific installation step details not mentioned in the materials are not yet provided.

Related tools

Browse category