Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

GitHub - secondlaw-ai/skyrim: 🌎 🤝 AI weather models united

May 07, 2024 - github.com
Skyrim is a platform that allows users to run large weather models in less than two minutes using a consumer-grade GPU. It aims to make weather forecasting more accessible by providing a well-maintained infrastructure that surpasses the skill level of traditional numerical models run on CPU HPC clusters. Users can simulate extreme weather events, fine-tune forecasts, and even run ensemble models. The platform supports various weather models and initial conditions, and it offers different methods for running forecasts, including using Modal or your own GPUs.

The platform provides detailed instructions for installation, running your first forecast, and setting up different environments. It also offers examples of how to run forecasts with different models, initial conditions, and lead times, and how to store and read forecasts in AWS. The roadmap for Skyrim includes ensemble prediction, an interface for real-time NWP-based predictions, a global model performance comparison, a fine-tuning API, and model quantization. The platform is built on top of NVIDIA's earth2mip and ECMWF's ai-models.

Key takeaways:

  • Skyrim is a platform that allows running large weather models in less than 2 minutes using a consumer-grade GPU. It aims to make these models accessible by providing a well-maintained infrastructure.
  • The platform supports running on modal, on a container, or bare metal. It also provides the option to use your own GPUs for forecasting.
  • Initial conditions for the weather models can be pulled from GFS, ECMWF IFS (Operational), or CDS (ERA5 Reanalysis Dataset). The platform supports various large weather models including Graphcast, Pangu, Fourcastnet, DLWP, Fuxi, and MetNet-3.
  • The roadmap for Skyrim includes ensemble prediction, interface to fetch real-time NWP-based predictions, global model performance comparison, finetuning API, and model quantization.
View Full Article

Comments (0)

Be the first to comment!