A Huggingface Application based on Gradio for Attracting Contributors and Fast Iteration

[RaccoonOnion]

Right now MOT has issues with validating data and there are couple issues with the code base as well. The obscurity of Drupal (used in the current codebase) hinders the fast iteration and bug tracing. It also increases the learning cost for new contributors. I think a Gradio based Huggingface app could be one option to change this situation:

1. Gradio is a an open-source Python package that allows fast development of a web application. The codebase is very simple compared with traditional packages. This would greatly accelarate the development and future maintanence. The app developed with Gradio can be hosted in Huggingface spaces easily too.
2. The big issue with data validation has to be done with community effort because a small number of core contributors can not keep track of all newly released models. Huggingface is currently the to-go places for people interested in using LLMs. Hosting in Huggingface could increase the exposesure and contribution for MOT.

If this idea were to be pursued, I would the following roadmap could work efficiently:

1. Generate a short list of most popular models now and validate the data manually
2. Develop a simple, intuitive app using Gradio and host it in Huggingface with the data acquired from step 1
3. Gather community suggestions and new data entries for continual improvement
4. • Use data and experience we gain from step 1-3 to improve the website codebase in Github

This is still a premature idea and look forward for more discussions!