metadata
pipeline_tag: image-to-image
library_name: pytorch
license: apache-2.0
tags:
- weather-radar
- precipitation-nowcasting
- diffusion-models
- generative-ai
- deep-learning
datasets:
- maly-javornik-shmu
Pretrained Checkpoints for Radar Precipitation Nowcasting Benchmark
Repository ini menyediakan bobot model (checkpoints) terlatih untuk inferensi dan evaluasi nowcasting presipitasi berbasis data radar SHMU Maly Javornik (domain 340x340 km2).
Arsitektur Model
| Model | Checkpoint File | Ukuran | Arsitektur Dasar |
|---|---|---|---|
| GenLUPIN | checkpoints/genlupin.ckpt |
~52 MB | Lagrangian-Eulerian Score-based EDM Diffusion (~4.29 M parameter) |
| LUPIN | checkpoints/lupin.ckpt |
~382 MB | Deterministic Dual U-Net with physics-based advection (~31.87 M parameter) |
| DiffCast | checkpoints/diffcast_simvp.pt |
~922 MB | Latent Diffusion Model based on SimVP and ContextNet (~57.56 M parameter) |
| MFUNet | checkpoints/mfunet.ckpt |
~2 MB | Motion Field U-Net untuk pemodelan temporal (~0.49 M parameter) |
Struktur Repositori
nowcasting-radar-checkpoints/ ├── README.md ├── .gitattributes └── checkpoints/ ├── genlupin.ckpt ├── lupin.ckpt ├── diffcast_simvp.pt └── mfunet.ckpt