GPUStack 0.1.2: Enhanced Experience for Deploying Models from Hugging Face

GPUstack 0.1.2

GPUStack 0.1.2 has been released. This update introduces enhancements and fixes inspired by community feedback to better support a variety of user scenarios. Key features include an enhanced experience for deploying Hugging Face models, with more detailed model information available during search and selection. Additionally, GPUStack now supports the use of Docker Registry as a private model repository.

 

For more details about GPUStack, visit:

GitHub repo: https://github.com/gpustack/gpustack

User guide: https://docs.gpustack.ai

 

What's New in GPUStack 0.1.2

Enhanced Experience for Deploying Models from Hugging Face

We’ve improved the model deployment experience from the Hugging Face by offering a more visual interface. Users can now see additional model information—such as trending, downloads, likes, quantization, and README—helping them make more informed model selections.

image-20240902175217446

 

Support for Docker Registry as a Private Model Repository in GPUStack

GPUStack now supports using Docker Registry as a private model repository, eliminating the need to download models from public repositories.

 

Recursive Search for Hugging Face Models

We’ve addressed an issue reported by the community where models stored in subdirectories of the Hugging Face repository were not deployable. With this fix, recursive searches are now supported for models in Hugging Face.

 

Bug Fixes in GPUStack 0.1.2

Fixes for Open WebUI Integration

Resolved issues with missing API response data when integrating Open WebUI with GPUStack, enhancing compatibility with third-party integrations.

 

Error Handling When No GPU is Present

Addressed issues where the frontend and backend would throw errors in environments without a GPU. These errors have now been fixed.

 

GPU Information Collection Improvements

Fixed an issue where GPU information was missing when both integrated and discrete GPUs were enabled simultaneously and their indexes overlapped.

 

Configuration File Parameter Fixes

Resolved issues where specific configuration parameters were ineffective when starting GPUStack using a configuration file.

 

Compatibility Fixes for Windows / CentOS (with SELinux enabled) / Ubuntu 24.04

We’ve addressed several installation and usage issues reported by the community, ensuring better compatibility with these operating systems.

 

User Information Sync Issue on Login Page

Fixed a bug where, after a session timeout requiring re-login, opening multiple GPUStack windows could repeatedly cause the interface to switch between the login page and the primary interface.

 

Join Our Community

You can see the release's full changelog here. As always, we'd love to hear your feedback via Discord or GitHub. Your feedback is invaluable in helping us make GPUStack better for everyone, and we welcome your contributions to the project.

Related Articles