NVIDIAOfficial·6 Agent Skills Included

nvalchemi-toolkit

GPU molecular dynamics, model training, and multi-GPU scaling

Runs batched molecular dynamics, geometry relaxation, and machine-learned potential training on NVIDIA GPUs. Guides wrapping custom models, composing multi-stage simulation pipelines, and splitting large systems across GPUs. Eliminates manual batching, checkpoint wiring, and distributed setup so simulations scale from one GPU to many.

127|32

↗

npx skills add NVIDIA/nvalchemi-toolkit --all -g -y

Available:

Tells the coding agent how to set up CUDA environments, follow repository conventions, and route tasks to the matching skill guides for dynamics, training, data, and distributed workflows.

All Skills in This Repository (6)

Pure Emerald Level Indicators

📦 In Repo

NVIDIA

nvalchemi-model-wrapping

Wrap PyTorch MLIPs with BaseModelMixin to standardize inputs, outputs, and embeddings.

nvalchemi-dynamics-implementation

Define custom dynamics integrators by overriding pre_update() and post_update() in BaseDynamics for GPU-accelerated atomic simulations.

nvalchemi-dynamics-api

Configure GPU-accelerated dynamics pipelines with FusedStage and DistributedPipeline.

nvalchemi-data-storage

Persist and load AtomicData with a Zarr-backed Writer, Reader, Dataset, and DataLoader pipeline.

nvalchemi-dynamics-hooks

Integrate hook-based callbacks into NVIDIA nvalchemi dynamics simulations.

nvalchemi-data-structures

Construct AtomicData and Batch objects for GPU-ready atomic graph workflows.

Official

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install nvalchemi-toolkit?▼

Run `npx skills add NVIDIA/nvalchemi-toolkit --all -g -y` in your terminal to install all skills in this suite globally.

How to run molecular dynamics on multiple GPUs?▼

The nvalchemi-distributed skill shows how to wrap any integrator in DomainParallel and launch with torchrun to split one large system across GPUs.

Can I use my own machine-learned potential model?▼

Yes. The model-wrapping skill explains how to adapt any PyTorch model with BaseModelMixin so it works with dynamics, training, and fine-tuning stages.

How to fine-tune a pretrained MACE or AIMNet2 model?▼

The fine-tuning skill covers FineTuningStrategy, pretrained checkpoint loading, parameter freezing, and conservative optimizer defaults for adapting models to new data.

Does nvalchemi-toolkit work with Claude Code and other agents?▼

Yes. All guides follow the universal SKILL.md standard and load automatically in Claude Code, Cursor, OpenCode, and Codex.