Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/rparrett/kilter_brain_gen
Kilter Board climb generation experiments
https://github.com/rparrett/kilter_brain_gen
Last synced: 18 days ago
JSON representation
Kilter Board climb generation experiments
- Host: GitHub
- URL: https://github.com/rparrett/kilter_brain_gen
- Owner: rparrett
- Created: 2024-03-25T02:36:42.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2024-08-29T16:07:09.000Z (3 months ago)
- Last Synced: 2024-10-15T08:48:08.991Z (30 days ago)
- Language: Python
- Size: 59.6 KB
- Stars: 0
- Watchers: 1
- Forks: 1
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# kilter_brain_gen
Generating kilter board problems with transformers
## Usage
### Clone the repo
```bash
git clone https://github.com/rparrett/kilter_brain_gen
cd kilter_brain_gen
```### Get the kilter sqlite database
- Install [`sqlite`](https://www.sqlite.org/download.html)
- Use [boardlib](https://github.com/lemeryfertitta/BoardLib) or extract from a kilter `apk` file.
- Run `get_csv.sh`### Install dependencies
- Install [just](https://github.com/casey/just?tab=readme-ov-file#installation)
- Run `just venv`
- Run `just sync-deps`
- Windows Only: Run `just torch-cuda`### Train the `climb` model
- `just run src/climb_clm/train_tokenizer.py`
- `just run src/climb_clm/train.py`### Generate some climbs
- `just run src/climb_clm/generate.py`
### Run the API server
- `just flask`
- Debug builds of [`kilter_brain`](https://github.com/rparrett/kilter_brain) will connect to the local server.## TODO
- [ ] Train a model for route names that actually works
- [ ] Tidy everything up with a nice CLI framework
- [ ] Add similarity search
- [ ] Somehow specify windows-specific dependencies that grab torch with CUDA support### Experiments
- [ ] Try adding duplicate climbs with randomized frame data ([shuffle-frames](https://github.com/rparrett/kilter_brain_gen/tree/shuffle-frames))
- [ ] Try different sampling strategies for climb data