Orchestrate multi-GPU Open Dreamer world-model training on Nebius clusters.
npx skills add https://github.com/nebius/nebius-physical-ai --skill open-dreamer
Please help me install this Agent Skill. Skill: open-dreamer Source: https://github.com/nebius/nebius-physical-ai/tree/main/skills/tools/open-dreamer Command: npx skills add https://github.com/nebius/nebius-physical-ai --skill open-dreamer
This skill addresses the complexity of onboarding and executing high-performance, multi-GPU world-model training pipelines, specifically the Dreamer 4 architecture, on Nebius infrastructure.
Execute the open-dreamer skill to initiate a multi-GPU training smoke test on the specified repository reference.
High-intent search queries and answers about installing and using this skill.
To train a Dreamer 4 world-model on multi-GPU infrastructure, you can use this skill to orchestrate JAX and Flax NNX data-parallel training meshes on Nebius clusters, handling causal video tokenization and latent dynamics.
Action-conditioned latent dynamics training for physical AI is a predictive modeling process that simulates world-model environments, enabling causal video tokenization and rollout visualization using tools like the Rerun viewer.
Yes, you need JAX and Flax NNX configured in your environment to run the Open Dreamer pipeline, along with a multi-GPU resource profile to manage data-parallel training meshes on Nebius infrastructure.
Yes, you can use causal video tokenization with Minecraft VPT datasets for robotics research to train the Dreamer 4 world-model and visualize predicted rollouts in the Rerun viewer.
FVD scoring in world-model simulation pipelines evaluates the quality of generated video predictions by calculating the Frechet Video Distance, which is facilitated by the skill during the action-conditioned latent dynamics training process.
Running multi-GPU data-parallel training meshes for world-models requires a configured Nebius multi-GPU resource profile and specific dependencies including JAX, Flax NNX, decord, and rerun-sdk to execute the pipeline successfully.
yonesukeBuild, train, and manage Flax NNX neural networks with state handling.
nebiusStandardize containerization and multi-GPU deployment of action-conditioned world models.
wangyendtAdds 3D visualization primitives to a Rerun viewer using Pywayne utilities.
elodin-sysCreate and modify JAX-based physics simulations for aerospace systems.
