WE-Dream Pretraining Models
Collection
Action-free, hand-pose, and latent-action pretrained models for robot post-training. β’ 5 items β’ Updated
This WE-Dream checkpoint was pretrained for 50,000 steps on approximately 5,000 hours of EgoDex, EgoVerse, and EgoWE egocentric video, supervised by 32-dimensional latent actions. It initializes robot post-training.
| Property | Value |
|---|---|
| Prediction horizon | 24 steps |
| Routed-attention probability | 0.5 |
Follow the installation instructions to set up the training environment and download the base model files. Accept the access terms on this page, then download:
hf download we-dream/pretrain-latent-5kh \
--local-dir checkpoints/we-dream-latent-5kh
Set PRETRAINED_MODEL_PATH=checkpoints/we-dream-latent-5kh when following the post-training guide.
Apache-2.0. Source revisions and file checksums are recorded in provenance.json.