hello_numerai.py
Train a small Numerai model and export hello_numerai.pkl
Python 3.11 and tried in python 3.12 both got the same error
pip install lightgbm==4.0.0 pandas==2.1.1 numpy==1.26.4 joblib==1.3.2 pyarrow==14.0.2 numerapi
pip -3.11 install lightgbm==4.0.0 pandas==2.1.1 numpy==1.26.4 joblib==1.3.2 pyarrow==14.0.2 numerapi
import joblib
import pandas as pd
import lightgbm as lgb
from numerapi import NumerAPI
def main():
print("📥 Downloading Numerai v5 training data...")
napi = NumerAPI()
napi.download_dataset("v5.0/train.parquet", "train.parquet")
print("📂 Loading data...")
train = pd.read_parquet("train.parquet")
# ✅ Select all feature columns (v5 uses long random names)
features = [c for c in train.columns if c.startswith("feature_")]
print(f"Found {len(features)} features")
# ✅ To fit on small PC: use only first 50 eras
eras = train["era"].unique()[:50]
train_small = train[train["era"].isin(eras)]
print(f"Training on {len(train_small)} rows from 50 eras")
X = train_small[features]
y = train_small["target"]
print("🤖 Training LightGBM model...")
model = lgb.LGBMRegressor(
n_estimators=50, # fewer trees
learning_rate=0.1, # faster training
max_depth=4, # shallow trees
num_leaves=15, # smaller leaves
max_bin=64, # reduce memory
n_jobs=-1,
random_state=42 # model seed
)
model.fit(X, y)
# ✅ Save model as hello_numerai.pkl
joblib.dump(model, "hello_numerai.pkl")
print("✅ Model saved as hello_numerai.pkl (upload this to Numerai)")
if name == "main":
main()
This the error I obtained/e-mail:
Model Upload Failed
Timestamp 2025-09-13 18:33 UTC
Status invalid
Python version numerai_predict_py_3_11
Data version v5.0
⚠️ Model failed to generate a live output!