rel-hm: item sales (Regression)
This runner now materializes rel-hm through relbench before building the GraphReduce frames. It no longer downloads parquet files from the example S3 bucket.
from relbench_hm_item_sales import run_rel_hm_item_sales
df_eval, df_holdout, holdout_mae, n_features, materialized, target = run_rel_hm_item_sales()
print(materialized)
print(target, holdout_mae, n_features, len(df_eval), len(df_holdout))
Current implementation:
Current RelBench-aligned dates:
- eval cut date:
2020-09-07 - holdout cut date:
2020-09-14 - lookback start:
2019-09-07
Latest Tested Result
Latest local run from tests/data/relbench/run_reports/relbench_results.json:
- test MAE:
0.0362 - test RMSE:
0.3638 - feature count:
35 - rows: train
5488184, validation105542, test105542 - runtime:
502.2s(2026-04-09T22:51:12Z)
Interactive Runner
Idle