This repository contain parts of the code to generate the 3rd place solution to the Kaggle's LLM Prompt Recovery competition. You can find the details for the solution in the write-up post
This repo contains the training scripts for the gate, clustering and tags models.
The two dataset files dataset.csv (for gate and cluster models) and the dataset_tags.csv (for the tag model) are also available here as a Kaggle dataset.
Before running the training scripts just make sure to create a new environment from the requirements.txt file, download the datasets from the provided link and unzip them here in the root directory.