earth2studio-create-prognostic

Create Earth2Studio prognostic model wrappers for time-stepping weather forecasts.

3.2k|370|Updated Feb 25, 2026
One-click install
npx skills add https://github.com/NVIDIA/skills --skill earth2studio-create-prognostic
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: earth2studio-create-prognostic
Source: https://github.com/NVIDIA/skills/tree/main/skills/earth2studio-create-prognostic
Command: npx skills add https://github.com/NVIDIA/skills --skill earth2studio-create-prognostic

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires torch, numpy, earth2studio, loguru, pytest, and includes references (resource) components.

What problem does it solve?

Integrating third-party ML weather models into Earth2Studio requires following strict conventions for coordinate systems, inheritance, testing, and documentation, which is error-prone without guidance.

Core Features & Use Cases

  • Structured Wrapper Creation: Generates prognostic model files with required triple inheritance (torch.nn.Module, AutoModelMixin, PrognosticMixin), canonical method ordering, and SPDX headers.
  • Test Generation: Produces mock and package integration tests following the Phoo dummy-model pattern, runnable via uv run pytest.
  • Docs and PR Workflow: Guides dependency extras in pyproject.toml, install docs, changelog updates, validation scripts, and PR templates.
  • Use Case: Ask your agent to add a Pangu-Weather wrapper from a reference inference script, and it creates the model file, tests, docs, and changelog entries following Earth2Studio conventions.

Quick Start

Ask your agent to create an Earth2Studio prognostic model wrapper for your weather model, optionally providing a reference inference script URL or path.

Frequently Asked Questions about earth2studio-create-prognostic

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create an Earth2Studio prognostic model wrapper?

Provide a reference inference script URL or path, then follow the guided workflow: analyze dependencies, create the model file under earth2studio/models/px with triple inheritance, write mock tests, and update docs and changelog. Run tests with uv run pytest.

What inheritance is required for Earth2Studio prognostic models?

Every prognostic model must inherit from torch.nn.Module, AutoModelMixin, and PrognosticMixin. Methods must follow canonical ordering from __init__ through input_coords, output_coords, load_model, __call__, and create_iterator.

Does this skill work for diagnostic models or data sources?

No, it is explicitly scoped to prognostic (time-stepping forecast) models only. Diagnostic models, data sources, and installation tasks are out of scope and should use other approaches.

Why does pytest fail with ModuleNotFoundError in Earth2Studio?

The environment requires running Python through uv to resolve dependencies. Use uv run pytest instead of plain pytest, and uv run python instead of python for all commands.

What latitude convention do Earth2Studio models use?

Public coordinates must use north-to-south latitude (90 to -90) and 0-to-360 longitude. If a source checkpoint expects south-to-north, flip tensors internally without exposing the flipped order in input_coords or output_coords.