snupilab/theta-bench-groot-sim-3003

THETA Bench GR00T N1.7: Simulation training, 3,003 segments

The validated final policy checkpoint is available in this repository.

This is a THETA training-result repository. It does not substitute an upstream pretrained policy for a THETA-trained checkpoint.

Setting	Value
Training stage	Simulation training, 3,003 segments
Target optimizer updates	40000
Per-GPU batch / GPUs / global batch	16 / 8 / 128
Gradient accumulation	1
Conditions per global batch	18
Dataset revision	8b2cd31e107b64cb13f812ea217a63a20845c78a

Pinned training data.

The simulation pool contains 1,200 successful L1/L2 demonstrations and 1,803 extracted L0 prefixes, spanning 18 conditions. The 3,003 segments are not 3,003 independent demonstrations.

Use the model's native THETA adapter and model-specific dependencies. This repository does not claim compatibility with arbitrary Transformers or simulation loaders. No evaluation score is claimed by checkpoint publication.

Load this directory with the native NVIDIA GR00T policy and the THETA simulation modality adapter (36-dimensional actions). The native loader first loads the backbone weights and processor from nvidia/Cosmos-Reason2-2B, pinned to revision 9ce19a195e423419c349abfc86fd07178b230561, before restoring this trained state. That upstream snapshot must be available separately. Use the pinned THETA/GR00T environment; this is not a generic Transformers AutoModel checkpoint.

Training uses independent model optimizers and shared GPU execution through MPS. Publication is performed by a CPU uploader after final checkpoint validation.