Numerai tournament example scripts using pytorch NN, lightGBM and optuna

https://numer.ai/tournament

Performance of my model based on this example

numerai model page
numerai model page

• 2022/04/25 Faster MLP model (nn.ModuleList -> nn.Sequential) & adopt numerai V4 data

• era-boosted train, time-series cross-validation
• era-batches training
• model hyperparameter tuning on pytorch NN and GBDT model
• several tips on Numerai Forum are also included

python3
gpu environment for pytorch # if you use pytorch NN model
virtualenv

git clone https://github.com/meaten/numerai_NN_example.git
cd numerai_NN_example
mkdir env
virtualenv env -p python3
source env/bin/activate
pip install -r requirements.txt

train model by era-boosted training. you can choose other config files also.

python src/main.py --config_file config/mlp.yml --gpu GPU_ID

test model for diagnostic.

python src/main.py --config_file config/mlp.yml --mode test --gpu GPU_ID

inference & submit.
Please specify follows.

• pairs of your model name and config file in src/main.py.
• Numerai API user id and secret key in src/default_param.py

python src/main.py --mode submit --gpu GPU_ID

tune model hyperparameter by optuna To train with tuned parameters, add LOAD_TUNED: True to the config file.

python src/main.py --config_file config/mlp.yml --mode tune --gpu GPU_ID

MIT

Please send me the bug report or wanted features on GitHub Issue or Numerai Forum.

If you find this repository helpful and feel generous, Please send NMR to my wallet address below.

0x0000000000000000000000000000000000025769