katha-ai/RecapStorySumm-CVPR2024 ? reverse-engineered prompt

Build me a working Python project for the paper “Previously On … From Recaps to Story Summarization” that can train and run inference on the PlotSnap dataset.

I want a setup where I can point it at my local data folder, load the provided pre extracted visual and dialog features, train the TaleSumm model, and then generate importance scores or summaries for episodes. Please include a simple way to configure paths, splits, checkpoints, and training settings, plus an inference script that loads a saved model and runs on an episode end to end.

If anything in the repo is unclear, follow the README and look up any current docs online if you need to. Make it easy for me to install, train, and test without having to figure out the whole research codebase first.