mjf-sumixed-apache-2.0-and-nvidia-open-model-license
cp_coal LIBERO-90 data bundle (Cosmos-Policy coalition training)
LoRA fine-tuning dataset for NVIDIA Cosmos-Policy-LIBERO-Predict2-2B on LIBERO-90 with coalition-averaged action loss, containing 42,809 VAE-encoded training examples and regenerated demonstrations.
Downloads19
Episodes403
Why This Matters for Physical AI
This dataset enables training of vision-language policy models for robotic manipulation through large-scale pretraining and fine-tuning on standardized benchmark tasks.
Technical Profile
- Modalities
- rgblanguage
- Environment
- simulation
- Task Types
- manipulation
- Episodes
- 403
- Data Format
- WebDataset
- Annotation Types
- language_instructionsaction_labels
- License
- mixed-apache-2.0-and-nvidia-open-model-license
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack