Two config layers: global (this directory) and per-project (in each projects/<name>/config/).
| File | Purpose | Tracked? |
|---|---|---|
alphapy.yml |
Local paths (data dir, project root). Read at startup by alphapy_main.get_alphapy_config(). |
gitignored |
alphapy.yml.template |
Template — copy to alphapy.yml and edit. |
tracked |
algos.yml |
ML algorithm definitions and hyperparameter grids (sklearn, XGBoost, CatBoost, LightGBM). Used by optimize.py. |
tracked |
model.yml.template |
Per-project model config template — copy to projects/<name>/config/model.yml. |
tracked |
Each project under projects/<name>/ has config/model.yml defining: target column, algorithms to run, CV folds, hyperparameter optimization settings, encoding, scaling, and feature selection. See projects/kaggle/config/model.yml for a working example.
cp alphapy.yml.template alphapy.yml
# Edit alphapy.yml with your data_dir and project_root pathsThen create a new project:
mkdir -p projects/myproject/{config,data}
cp config/model.yml.template projects/myproject/config/model.yml
# Edit model.yml: set target column, algorithms, etc.
cd projects/myproject && alphapy- Trading/markets/finance configs (
sources.yml,systems.yml,groups.yml,indicators.yml, finance-flavoredvariables.yml) moved to the privatealphapy-financerepo in v4.0.0. algos.ymlis shared and domain-agnostic. All entries are pure ML algorithms.