Upload AirQualityBench.py with huggingface_hub

AirQualityBench.py

@@ -0,0 +1,95 @@
+"""AirQualityBench dataset loading script for Hugging Face datasets."""
+
+import os
+import h5py
+import numpy as np
+import datasets
+
+
+_CITATION = """@article{airqualitybench2025,
+  title={AirQualityBench: A Global-Scale Air Quality Forecasting Benchmark
+         for Spatio-Temporal Graph Neural Networks},
+  author={...},
+  journal={arXiv preprint arXiv:2605.05854},
+  year={2025}
+}"""
+
+_DESCRIPTION = """AirQualityBench is a global-scale air quality forecasting benchmark
+featuring 3,720 monitoring stations across the world with **authentic missing patterns**
+and physical-scale evaluation. The dataset spans 5 years (2021–2025) of hourly
+measurements across 6 primary pollutants: PM2.5, PM10, NO2, O3, SO2, CO.
+"""
+
+_HOMEPAGE = "https://github.com/Star-Learning/AirQualityBench"
+_LICENSE = "MIT"
+
+NUM_STATIONS = 3720
+NUM_POLLUTANTS = 6
+
+
+class AirQualityBench(datasets.GeneratorBasedBuilder):
+    VERSION = datasets.Version("1.0.0")
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(name="train", description="Training set (2021–2023)"),
+        datasets.BuilderConfig(name="validation", description="Validation set (2024)"),
+        datasets.BuilderConfig(name="test", description="Test set (2025)"),
+        datasets.BuilderConfig(name="2021", description="Year 2021 hourly data"),
+        datasets.BuilderConfig(name="2022", description="Year 2022 hourly data"),
+        datasets.BuilderConfig(name="2023", description="Year 2023 hourly data"),
+        datasets.BuilderConfig(name="2024", description="Year 2024 hourly data"),
+        datasets.BuilderConfig(name="2025", description="Year 2025 hourly data"),
+    ]
+
+    DEFAULT_CONFIG_NAME = "train"
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features({
+                "timestamp_idx": datasets.Value("int32"),
+                "values": datasets.Array2D(
+                    shape=(NUM_STATIONS, NUM_POLLUTANTS), dtype="float32"
+                ),
+                "masks": datasets.Array2D(
+                    shape=(NUM_STATIONS, NUM_POLLUTANTS), dtype="int8"
+                ),
+            }),
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = os.path.dirname(__file__) or "."
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"data_dir": data_dir, "config": self.config.name},
+            )
+        ]
+
+    def _generate_examples(self, data_dir, config):
+        config_year_map = {
+            "2021": [2021], "2022": [2022], "2023": [2023],
+            "2024": [2024], "2025": [2025],
+            "train": [2021, 2022, 2023],
+            "validation": [2024],
+            "test": [2025],
+        }
+        years = config_year_map[config]
+
+        global_idx = 0
+        for year in years:
+            file_path = os.path.join(data_dir, f"aq_compact_{year}.h5")
+            with h5py.File(file_path, "r") as f:
+                values = f["values"][:]
+                masks = f["masks"][:]
+
+            for t in range(values.shape[0]):
+                yield global_idx, {
+                    "timestamp_idx": t,
+                    "values": values[t].astype(np.float32),
+                    "masks": masks[t].astype(np.int8),
+                }
+                global_idx += 1