Downloads · 30 days
0
justinchuby/pangu-weather-1h-onnx-catalogue
pangu-weather-1h-onnx-catalogue is a time series forecasting model from justinchuby. Use it for the time series forecasting task on the model card, and read the license before you ship it in a product. The card lists the license as cc-by-nc-sa-4.0.
Real published Pangu-Weather 1-hour checkpoint, packaged without weight modification. The model is stateless and maps global 0.25° upper-air and surface fields to a one-hour forecast.
Downloads · 30 days
0
Access
Public
Updated Aug 26, 2026
Repo size
1.4 GB
Likes
0
Public
Click a slice to open those files.
.onnx1.2 GB · 85%
From the Hugging Face model README
Real published Pangu-Weather 1-hour checkpoint, packaged without weight modification. The model is stateless and maps global 0.25° upper-air and surface fields to a one-hour forecast.
pangu_weather_1.onnx179e5029c453ae459dfd52f14610a6c5f5ad39f1371985744ea0ce6c546fda2a198808xc/Pangu-Weather@72bdd99096721e1a1f8912c37a9a3aff9ff0a4f2input: float32 [5,13,721,1440], variables Z/Q/T/U/V and pressure levels 1000..50 hPa.input_surface: float32 [4,721,1440], variables MSLP/U10/V10/T2M.output: float32 [5,13,721,1440], one-hour upper-air forecast.output_surface: float32 [4,721,1440], one-hour surface forecast.inference_metadata.yaml is the hashless canonical workflow-SSA contract. Graph hash and source integrity are recorded separately in graph_report.json and provenance.json.
hf download justinchuby/pangu-weather-1h-onnx-catalogue --repo-type model --local-dir pangu-weather-1h
cd pangu-weather-1h
sha256sum -c model.sha256
pip install 'onnxruntime-gpu==1.28.0' 'numpy==2.3.3'
python run.py --request request.npz --output reproduced_output.npz --provider cuda
The checked-in output.npz was produced from request.npz on NVIDIA H200 using CUDAExecutionProvider. The request is a deterministic physically scaled baseline, not an observed ERA5 analysis. Exact timing and sampled forecast values are in runtime_output.json.
Review inference_metadata.annotated.yaml for inline explanations of this package's workflow, tensor/state/cache contracts, and fail-closed omissions. inference_metadata.yaml remains the canonical machine-authored contract; automated validation confirms both files parse to the same metadata object.