learn it up

This project contains the (work in progress) code, and reference material for Learn it Up, a project that tries to adapt the LSTM and Transformer architectures for Pump it Up chart generation.

File structure

This project assumes that the songs packs are in ./data/songs, in usual packdir/songdir fashion. Step files must be in SSC.

The processing pipeline consists of:

• Parse SSC files
• Extract relevant keys from SSC (e.g DESCRIPTION, OFFSET, BPMS, NOTES) (Chart and StepFile) in code.
• Introduce mirroring chart variations (vertical only for S17 and below)
• Extract absolute time information for all charts
• Save processed ssc files to disk, at path ./data/parsed/${PACK_NAME}___${SONG_TITLE}.ready.ssc.bin (I wanted to use JSON but serializing and deserializing python classes to json sucks in general and i didnt want to add another dependency)

Running

The dependencies for this project are listed in pyproject.toml. This project is compatible with uv.

Use python3 main.py --help to get a list of available operations.

• python3 main.py parse_all to parse all SSC files.
• python3 main.py extract_all to extract features from all audio files.

Notebooks

The dataset analysis marimo notebook is in analysis.py, that can be run with marimo edit analysis.py or uv marimo edit analysis.py