This repository has various half-finished models and scripts for dealing with KilterBoard climbs (generating/grading/autoencoding).
conda create -n kilter -f env.yaml
Use the preprocess.ipynb
notebook to preprocess the data, generating the necessary csv files in the data/
directory.
Use the python diffuse.py
command to train a diffusion model for climb generation.
Use the python train.py
command to train a model to predict the difficulty of the climb.
Use the sample.ipynb
notebook to load trained diffusion and prediction models together to generate climbs.