Integrate AI Writing Feedback and Autocomplete with Transformers #47
Conversation
- Added AIFeedback.get_feedback() to generate detailed writing feedback using MBZUAI/LaMini-Flan-T5-248M. - Extended AIFeedback with get_autocomplete() (implementation pending) to generate writing suggestions. - Integration enables AI copilot-like behavior with inline suggestions. - Updated error handling for CUDA fallbacks. - Added transformers as a dependency in setup.py under install_requires. This update improves the writing experience by leveraging state-of-the-art NLP models.
|
Missing newline at end of file. How does this work when there is no internet connection? The application must not hang. What test have you used? |
I am using a 738 Million parameters model, it uses around 1 GB of GPU if available or like 10 seconds if on CPU. Not sure how it holds up on action OLPC laptop. Also this was done because I saw a summer project regarding this which was hard and I find really think it was hard.. we are planning to make sugar API first, which is the latest repo in my respositories now, and then use that instead. Rather than fully local. There's a docker image on that if you would like to try it. |
|
I mean what test of no internet connection. Also, any changes to dependencies; does something need to be installed first? README.md lists some, but you did not change it. |
Changes;
Note:
Video:
2025-02-04.15-59-18.mp4