zai-org

/

GLM-4.5V

like 581

Follow

Z.ai 4.06k

This PR significantly enhances the model card by integrating comprehensive details from the official GLM-V GitHub repository. The updates include:

• Updated Paper Link: The paper link now points directly to the Hugging Face paper page: GLM-4.1V-Thinking and GLM-4.5V: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning.
• Comprehensive Quick Links: Reorganized and expanded the top section to include direct links to the GitHub repository, online demo, API access, and the recently released desktop assistant app.
• Detailed Model Overview: Replaced the brief introduction with an in-depth "Model Overview" from the GitHub README, providing more context on GLM-4.5V's architecture, capabilities, and its relation to the GLM-V series, including relevant benchmark images.
• Project Updates: Added a dedicated section for "Project Updates" to keep users informed about recent developments and releases.
• Expanded Usage Instructions: Enhanced the "Quick Start" section with detailed environment installation steps, vLLM and SGLang integration, and specific examples for "Grounding" and "GUI Agent" to better showcase the model's functionalities.
• Model Fine-tuning Guide: Included a new section detailing how to fine-tune the model using LLaMA-Factory.
• Transparency on Issues: Added a "Fixed and Remaining Issues" section to provide clarity on known limitations and ongoing improvements.

These changes aim to make the model card a more complete and useful resource for the Hugging Face community.