Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 80, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

stormer-40yrs-canon — ERA5 1.40625° (properly downsampled)

Stormer 학습용 ERA5 h5 데이터셋. 1979–2020, 6시간 간격, 1.40625° (128×256).

⚠️ 기존 stormer-40yrs 와의 차이 — 반드시 읽을 것

기존 KyleBae1017/stormer-40yrs 는 WeatherBench2 의 240×121 (1.5°) 판을 1.40625° 로 업샘플해서 만들어졌다. 격자만 1.40625° 이고 내용물은 실제로 1.5° 였다.

이 데이터셋은 512×256 판에서 보존적으로 다운샘플한 정본이다.

stormer-40yrs (구) stormer-40yrs-canon (본)
소스 240x121_equiangular_with_poles_conservative (1.5°) 1959-2022-6h-512x256_equiangular_conservative
방향 업샘플 ⬆️ 다운샘플 ⬇️
lon.npy 인덱스 [0..255] (경도 아님) 실제 경도 [0, 1.40625, …, 358.59]

영향

구 데이터로 학습한 모델은 1.5° 왕복 오차가 성능 바닥으로 깔린다:

변수 왕복 바닥 24h RMSE 대비
T2m 0.520 K 68%
Z500 16.4 m²/s²
MSLP 27.4 Pa

추론 시점에 정본 초기조건을 넣어도 회수되지 않는다(모델이 이미 업샘플 분포를 학습했기 때문). 재학습이 유일한 경로다.

🚨 lon.npy 이중 보정 주의

구 데이터셋의 lon.npy 는 경도가 아니라 인덱스 [0..255] 였다. 이걸 좌표로 그대로 쓰면 conservative regrid 가 경도 방향으로 필드를 재배치해 결과 전체가 오염된다. 그 버그를 보정하던 코드가 있다면 이 데이터셋에서는 반드시 제거해야 한다 — 그대로 두면 이중 보정이 된다.

구성

wb2_h5df/
  train/  1979_0000.h5 … 2018_1459.h5   (40년, 58,440개)
  val/    2019_0000.h5 … 2019_1459.h5   (1,460개)
  test/   2020_0000.h5 … 2020_1463.h5   (1,464개)
  normalize_mean.npz, normalize_std.npz
  normalize_diff_{mean,std}_{6,12,18,24}.npz
  lat.npy, lon.npy

h5 하나 = 한 시각. f['input'][변수명] → float32 (128, 256). 82키 = 예보변수 69개 (지표 4 + 기압면 5×13) + 상수 12개 + time. 파일 하나는 정확히 10,646,557 바이트.

기압면 13개: 50, 100, 150, 200, 250, 300, 400, 500, 600, 700, 850, 925, 1000 hPa

받기

hf download KyleBae1017/stormer-40yrs-canon --repo-type dataset --local-dir .
cat wb2_h5df.tar.part-* | tar -xf -

검증

정본 재생성·검증 스크립트는 stormer 저장소regen_canon_data.sh / verify_regen_output.py 참조. 후자는 소스에서 표본을 다시 받아 배열 단위로 대조한다.

이 데이터셋의 val/test 는 공식 Stormer 체크포인트로 WeatherBench2 스코어카드를 평균편차 −0.05%, 평균|편차| 0.60% 로 재현하는 것이 확인됐다 (732 init, best-32, 50개 비교점).

Downloads last month
278