lizhuoq/WeatherLearn

python = "^3.11"
torch = ">=2.1.0"
timm = ">=0.9.10"
...

설치 (편집 가능 모드):

pip install -e .
# 또는 `poetry lock` / `poetry install` 후 Poetry 사용

단위 테스트 실행 (전체 Pangu()는 건너뜀 unless WEATHERLEARN_RUN_HEAVY=1):

python -m unittest discover -s tests -v

# Pangu
from weatherlearn.models import Pangu
import torch
...

# Pangu_lite
from weatherlearn.models import Pangu_lite
import torch
...

@article{bi2023accurate,
title={Accurate medium-range global weather forecasting with 3D neural networks},
author={Bi, Kaifeng and Xie, Lingxi and Zhang, Hengheng and Chen, Xin and Gu, Xiaotao and Tian, Qi},
...```

@article{bi2022pangu,
title={Pangu-Weather: A 3D High-Resolution Model for Fast and Accurate Global Weather Forecast},
author={Bi, Kaifeng and Xie, Lingxi and Zhang, Hengheng and Chen, Xin and Gu, Xiaotao and Tian, Qi},
...```

from weatherlearn.models import Fuxi
import torch
if __name__ == '__main__':
...

마지막 레이어는 논문과 일치합니다: Linear patch FC → 720×1440으로 reshape → 721×1440으로 bilinear interpolate. docs/fuxi_last_layer.md 참고

(Issue #8 명확화).

npj Climate and Atmospheric Science에 게재: FuXi: a cascade machine learning forecasting system for 15-day global weather forecast

by Lei Chen, Xiaohui Zhong, Feng Zhang, Yuan Cheng, Yinghui Xu, Yuan Qi, Hao Li

Multi-modal encode–fuse–decode (arXiv:2304.02948). 기본값은 13 levels / 69 channels (OpenEarthLab 실용적); 논문에서는 n_levels=37을 통해 37 levels / 189 channels 사용

from weatherlearn.models import FengWu, FengWu_lite
import torch
# Lite (smoke / small grid). ARCO 128x64 zarr는 Lat×Lon (64, 128)으로 전치됨.
...

연기(smoke) 학습 (공개 GCS ERA5): python examples/fengwu/train_smoke_gcs.py

• FengWu Model (https://arxiv.org/pdf/2304.02948v1.pdf) — weatherlearn/models/fengwu, examples/fengwu/ 참고

• FuXi Model (https://arxiv.org/pdf/2306.12873v3.pdf)

• Fuxi 모델에서 경도(longitude)와 위도(latitude)에 별도의 window_size를 설정하세요.

• 더 많은 유닛 테스트(unittest)를 추가하세요.

• 공식 Pangu 저장소에서 제공하는 사전 학습된 가중치(pre-trained weights)를 사용하여 Pangu 모델을 추론하세요.