Metadata-Version: 2.4
Name: sl-vla
Version: 0.0.1
Summary: VLA training framework.
Home-page: https://github.com/
Author: Lucas Jin
Author-email: jinfagang@163.com
License: GPL-3.0
Keywords: deep learning,script helper,tools
Platform: any
Classifier: Operating System :: OS Independent
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: Topic :: Education
Classifier: Topic :: Scientific/Engineering
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Scientific/Engineering :: Image Recognition
Classifier: License :: OSI Approved :: GNU General Public License v3 (GPLv3)
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.6
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Description-Content-Type: text/markdown
Dynamic: author
Dynamic: author-email
Dynamic: classifier
Dynamic: description
Dynamic: description-content-type
Dynamic: home-page
Dynamic: keywords
Dynamic: license
Dynamic: platform
Dynamic: summary

# Starlord

Dual system training framework for robotics.

The overall structure borrowed from GR00T, with these modifications:

- Simplifer VLM model, introduced more advanced VLM and larger slow system;
- Connecting fast system without cross-attention, using text encoder instead;



## Slow thought system

will goes `starlord/model/backbone` contains various VLMs. Provides a unified interface connect with fast system (action expert)



## Fast action system

Currently support:

- DiT: diffusion transformer
- QwenFlow: flowmatching based action expert



## Training

training scripts goes to `scripts/xxx.sh`

