WorldKernels
GPU-first world model simulation engine — serve learned world models (DiT, VAE) as interactive sessions.
⚠️ Early Development — This package is in pre-alpha. APIs may change.
Installation
pip install worldkernels
For full functionality:
pip install worldkernels[all] # Everything
pip install worldkernels[serve] # HTTP/WebSocket server
pip install worldkernels[diffusers] # HuggingFace Diffusers support
Quick Start
from worldkernels import WorldKernel, Action, WorldConfig
# Initialize engine
wk = WorldKernel(device="cuda")
# Load a world model from HuggingFace Hub
wk.load_world("Etched/oasis-500m")
# Create an interactive session
session = wk.create_session(
world="oasis-500m",
config=WorldConfig(height=360, width=640, fps=20),
)
# Step through the simulation
for _ in range(100):
action = Action("keyboard", {"keys": ["W", "SPACE"]})
obs = session.step(action)
# obs.frames contains generated video frames
session.close()
wk.shutdown()
Features (Planned)
- 🎮 Session-based API — Stateful simulation with checkpoint/branch
- 🚀 GPU-optimized — Pre-allocated buffers, CUDA graphs, torch.compile
- 🔌 HuggingFace native — Load models directly from the Hub
- 🌐 HTTP/WebSocket server — REST API and real-time streaming
- 🧩 Extensible backends — PyTorch eager, torch.compile, TensorRT
Documentation
Coming soon at worldkernels.dev
License
LGPL-2.1 — see LICENSE
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
worldkernels-0.1.0.tar.gz
(31.8 kB
view details)
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file worldkernels-0.1.0.tar.gz.
File metadata
- Download URL: worldkernels-0.1.0.tar.gz
- Upload date:
- Size: 31.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.2.0 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
eccdf68439246eb4102fda9b80082d37790fada7dc18a9bbe528041b50d0bcb3
|
|
| MD5 |
4055f3c3d279c2537b5fda7834330b01
|
|
| BLAKE2b-256 |
d75b922110c682b1a61cf598e3f5c4f6abfb8c98a4c60f719aeb7240603e6f67
|
File details
Details for the file worldkernels-0.1.0-py3-none-any.whl.
File metadata
- Download URL: worldkernels-0.1.0-py3-none-any.whl
- Upload date:
- Size: 31.7 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.2.0 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
925103281f5d758027f8e31f80ac233311851419b55ada82ba1cbb936bb00941
|
|
| MD5 |
41ac5e0725ff7c36f481b379e6363554
|
|
| BLAKE2b-256 |
cca44e716f58e8ca3b6d302f765a2ec6c2a1952155bfab3eb363d8c7e5be44bc
|
