Скачать с ютуб Look at Your Data: Building an LM Scoring App with FastHTML в хорошем качестве

Look at Your Data: Building an LM Scoring App with FastHTML

In this video, I walk through a scoring app I created to evaluate text generations from different language models. I built this app using FastHTML (with a lot of help from Claude!) to implement my scoring rubric across six categories: grammar, creativity, plot, factual knowledge, reasoning, and context tracking. Starting with the basics, I progressively build functionality piece by piece, showing how I: Set up data loading from CSV files (criteria.csv, lmdata.csv, output.csv) Created a simple UI that displays prompts and completions Implemented category selection buttons Added scoring functionality with 0, 0.5, and 1.0 buttons Built navigation between different prompts Made sure scores persist when navigating between prompts/categories I'm not a FastHTML expert (as you'll see!), so I'm sure there are more elegant ways to implement some of this functionality. All code is available in my TinyScaleLab repo, including the scoring app and the data files needed to run it: https://github.com/vishalbakshi/TinyS... TinyScaleLabe playlist:    • TinyScaleLab