The Disappointing Reality of text-generation-webui: A Frustrating Journey Filled with Broken Promises and Subpar Results

[ZipingL]

I started using this project over a year ago when GPT suddenly gained popularity. I wanted to begin working with language models, but I needed a solid foundation to build upon.

Back then, I found it frustrating that there were limited resources available to run my own language model. Even now, the situation hasn't improved significantly. I experimented with several different projects, but none of them met my expectations. Either the instructions were unclear, or the models themselves were subpar. However, I stumbled upon this particular project that claimed to be like an automatic1111 for language models, which piqued my interest. I was hopeful that this might be the solution I had been seeking. Setting it up, however, proved to be a real challenge. I had to go through a lot of trouble to install the correct packages required by the webui, and I also had to install several additional libraries to enable other functions.

Despite the initial difficulties, I was eventually able to make it work. Fast forward to the present, and I am amazed at the progress this project has made. The seamless setup scripts and the web interface for automation have greatly enhanced the user experience, making it much more accessible and user-friendly.

I find this project to be incredibly helpful, especially since I don't have access to a powerful server or have specialized needs. Writing my own inferencing or training program would not provide me with significant advantages. At most, I can only run a 13B parameter model with decent speeds, which is sufficient for my current requirements. I want to express my gratitude to the developers and contributors of this project for their hard work and dedication.

Please keep up the excellent work, and continue to improve and expand this valuable resource for the community.

Have a great day!