NVIDIA NIM for Earth-2 FourCastNet# Earth-2 FourCastNet Overview References Advantages of NIMs Release Notes Release 2.0.0 Summary Key Features Notes and Limitations Prerequisites Support Matrix Model Support Hardware Support Optimized Configuration Non-optimized Configuration Software NGC Account Client Side Python Dependencies Next Steps Quickstart Guide Workflow Overview Launching the NIM Checking NIM Health Bash Python Fetching Input Data Inference Requests Bash Python Results Streaming Responses Post Processing Results Deployment Guide View Available NIM Container Information Pull the Container Image Docker NGC CLI Run the Container Checking the Status of the FourCastNet NIM Stop a Running Container Model Checkpoint Caching Air Gap Deployment (Offline Cache) Air Gap Deployment (Local Model) Configure NIM at Runtime GPU Selection Shared Memory Flag Model Profiles FourCastNet 3 ERA5 72ch Environment Variables Mounted Volumes Observability Performance Evaluation Process Performance Results H100 Replicating Benchmarking Results API Reference OpenAPI Schema Reference Troubleshooting Container and Deployment Issues Container Fails to Start Error Downloading Model Triton Model Failed to Launch (No GPU Visible) Container Starts But Health Check Fails Port Already in Use API and Inference Issues Request Timeout HTTP 400 or 500 Errors on /v1/infer Data Format Issues Input Array Shape or Dtype Mismatch Performance and Resource Issues Slow Requests Air Gap / Offline Deployment Issues Model Not Found in Offline Mode Getting Additional Help
