CoolFace
Modelpublic

Physis-AI/DreamDojo-G1-14B-Diffusers

sourceHugging Faceotherupdated 5mo agoView on Hugging Face
0likes
Model Card

DreamDojo-G1-14B-Diffusers

Large model fine-tuned on G1. Part of the DreamDojo model family.

Size14B
StagePost-training
ArchitectureDiT (Diffusion Transformer) with AdaLN-LoRA
BaseCosmos Predict 2.5

Checkpoint Structure

DreamDojo-G1-14B-Diffusers/
├── transformer/            # DiT backbone (sharded safetensors)
├── crossattn_adapter/      # Text-to-DiT projection (100352 → 1024)
├── vae/                    # AutoencoderKLWan (standard diffusers)
├── lam/                    # Latent Action Model (710M params)
├── text_encoder/           # Cosmos-Reason1-7B
├── scheduler/              # FlowMatchEulerDiscreteScheduler
├── action_processor/       # DreamDojo-specific config
└── config.json

Architecture

14B
Model channels5120
Transformer blocks36
Attention heads40
Patch size (spatial / temporal)2 / 1
Action dim384 (unified)

Citation

bibtex
@article{dreamdojo2025,
  title={DreamDojo: Advancing Real-World Robot Policies Through Generated Interactive Environments},
  author={NVIDIA},
  year={2025}
}

License

Please refer to the NVIDIA DreamDojo repository for license terms.