physicsnemo
Open-source deep-learning framework for building, training, and fine-tuning deep learning models using state-of-the-art Physics-ML methods
PhysicsNeMo is a PyTorch-based framework for physics machine learning that includes two agent skills for coding assistants. The PhysicsNeMo Discover skill helps agents find models, datapipes, examples, and documentation for scientific ML tasks, while the ShardTensor skill assists with implementing domain parallelism and multi-GPU operations. These skills enable AI coding agents to leverage PhysicsNeMo's extensive library of physics-informed deep learning models, from weather forecasting to structural mechanics, by providing contextual guidance and code generation support.
Key Features
Use Cases
- 01Helping coding agents discover appropriate neural operator models for PDE surrogate tasks
- 02Guiding AI assistants to implement domain parallelism for large-scale scientific simulations
- 03Assisting agents in selecting the right weather forecasting model from AFNO, GraphCast, or Pangu-Weather families
- 04Enabling automated generation of multi-GPU training scripts with proper DDP or FSDP2 integration
- 05Supporting agents in finding relevant examples for aerodynamics, structural mechanics, or healthcare applications
- 06Facilitating AI-assisted implementation of shard-aware operations for distributed scientific computing
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
physicsnemo — FAQ
What are the PhysicsNeMo agent skills?+
PhysicsNeMo includes two agent skills: PhysicsNeMo Discover helps coding agents find models, datapipes, examples, and documentation for scientific ML tasks, while PhysicsNeMo ShardTensor assists with implementing domain parallelism and multi-GPU operations. Both skills are designed for Codex, Claude Code, and compatible coding agents.
How do I install the PhysicsNeMo skills for my coding agent?+
The skills are located in the skills/ directory of the PhysicsNeMo repository at skills/physicsnemo-discover/SKILL.md and skills/physicsnemo-shard-tensor/SKILL.md. Place these skill files in your coding agent's skill directory according to your agent's documentation (location varies by agent platform).
Which coding agents support PhysicsNeMo skills?+
The skills work with Codex, Claude Code, and other compatible coding agents that support the skill integration format used by PhysicsNeMo. Each skill includes a benchmark evaluation demonstrating its effectiveness.
Do I need API keys or the PhysicsNeMo framework installed to use the skills?+
While the skills themselves are documentation files that guide the agent, to execute generated code you'll need PhysicsNeMo installed (pip install nvidia-physicsnemo). No additional API keys are required for the skills, though your coding agent itself may require authentication.
Are the PhysicsNeMo agent skills free to use?+
Yes, the skills are open-source and released under the Apache License 2.0 as part of the PhysicsNeMo project. They are free to use for both research and commercial applications.
What's the difference between the two PhysicsNeMo skills?+
PhysicsNeMo Discover focuses on helping agents navigate the framework's models, examples, and documentation to solve SciML tasks. PhysicsNeMo ShardTensor specializes in distributed computing aspects, helping agents implement domain parallelism, integrate DDP/FSDP2, and write multi-GPU code.
How do I install physicsnemo?+
Open the source repository on GitHub and follow its README. physicsnemo is a skill — MCP Agents Market links you directly to the official repo.
Is physicsnemo free?+
physicsnemo is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.