ArchEGraph/ArchEGraph

PACK

PACK is a building-energy dataset organized for graph-based and weather-conditioned learning.

Dataset Summary

• Total cases in manifest.csv: 49,326
• Unique buildings: 5,481
• Unique weather IDs: 64
• n_steps range: 968 to 8,760
• n_spaces range: 1 to 231

This repository currently stores:

• manifest.csv (index of all cases)
• building/ (5,481 files)
• geometry/ (5,482 files)
• weather/ (64 files)
• energy/ (49,326 files; nested under subfolders like 00/)
• split/ (predefined split CSV files)

Data Layout

Each row in manifest.csv contains:

• sample_id: case ID (building__weather style)
• source_job_tag: source identifier
• weather_id: weather/location key
• building_id: building key
• energy_file: relative path to energy npz file under energy/
• n_steps: number of time steps
• n_spaces: number of spaces/zones

Included Split Files

• split/split_p.csv (30,658 rows)
• split/split_m.csv (18,668 rows)
• split/split_demo.csv (300 rows)
• split/split_building_bias.csv (3,000 rows)
• split/split_weather_bias.csv (3,000 rows)

Each split CSV uses the same columns:

• case_id, sample_id, building_id, weather_id, subset, split, scenario

Quick Start

import pandas as pd
from pathlib import Path

root = Path(".")  # dataset root
manifest = pd.read_csv(root / "manifest.csv")

row = manifest.iloc[0]
energy_path = root / "energy" / row["energy_file"]
building_path = root / "building" / f"{row['building_id']}.npz"
weather_path = root / "weather" / f"{row['weather_id']}.npz"

print(row["sample_id"])
print(energy_path, building_path, weather_path)

Notes

• Energy files are stored in nested folders referenced by energy_file; do not assume all files are directly under energy/.
• If you need a smaller download, use a demo subset generated from split/split_demo.csv.

Citation

If you use this dataset, please cite your project/paper and this Hugging Face dataset page.