--- pretty_name: RIFT Checkpoints license: other license_name: model-specific-licenses license_link: https://huggingface.co/PoopBear/RIFT/blob/main/README.md#license library_name: pytorch base_model: - Wan-AI/Wan2.2-TI2V-5B - nvidia/Cosmos-Predict2-2B-Video2World datasets: - yuanty/LIBERO-fastwam - nvidia/LIBERO-Cosmos-Policy tags: - robotics - robot-learning - world-action-model - flow-matching - libero - robocoin - galaxea - arxiv:2608.11521 --- # RIFT Checkpoints Checkpoints for **[Keep the Future, Drop the Rollout: RIFT for World Action Models](https://arxiv.org/abs/2608.11521)**. [Code and evaluation](https://github.com/ChushanZhang/RIFT). ## Checkpoints Each dataset has one top-level directory. Model variants live inside that directory. | Dataset | Directory | Models | | --- | --- | --- | | LIBERO | [libero](https://huggingface.co/PoopBear/RIFT/tree/main/libero) | FastWAM; RIFT, FastWAM-IDM and FastWAM-Joint (step 21,700); Cosmos Policy Predict2-2B | | RoboCOIN, three tasks | [robocoin_multitask_mm](https://huggingface.co/PoopBear/RIFT/tree/main/robocoin_multitask_mm) | RIFT (final), 10 epochs / step 9,340 | | Galaxea indoor cleaning | [galaxea_indoor_cleaning_3cam224](https://huggingface.co/PoopBear/RIFT/tree/main/galaxea_indoor_cleaning_3cam224) | FastWAM, FastWAM-Joint and RIFT, 10 epochs / step 14,830 | Shared Wan VAE, text-encoder, and tokenizer files are under `assets/`. Exact checkpoint paths, hashes, and source revisions are listed in [checkpoint_manifest.json](checkpoint_manifest.json). Use each model's own normalization statistics. ## Download LIBERO Cosmos-2: ```bash hf download PoopBear/RIFT --include 'libero/cosmos-2/*' --local-dir ./checkpoints ``` LIBERO FastWAM: ```bash hf download PoopBear/RIFT --include 'libero/fastwam/*' --local-dir ./checkpoints ``` LIBERO RIFT: ```bash hf download PoopBear/RIFT --include 'libero/rift/*' --local-dir ./checkpoints ``` RoboCOIN RIFT: ```bash hf download PoopBear/RIFT --include 'robocoin_multitask_mm/rift_10ep/*' --local-dir ./checkpoints ``` Galaxea RIFT: ```bash hf download PoopBear/RIFT --include 'galaxea_indoor_cleaning_3cam224/rift_10ep/*' --local-dir ./checkpoints ``` The real-world model folders include saved training configurations. LIBERO model configs must be composed with the corresponding task and data configs. Set local dataset and pretrained asset paths for your installation. Load PyTorch checkpoints with `weights_only=True`. ## Citation ```bibtex @article{zhang2026rift, title={Keep the Future, Drop the Rollout: RIFT for World Action Models}, author={Zhang, Chushan and Tong, Jinguang and Li, Xuesong and Wang, Yikai and Li, Hongdong}, journal={arXiv preprint arXiv:2608.11521}, year={2026} } ``` ## License RIFT checkpoints are distributed under the Apache License 2.0 (see `LICENSE`). RIFT source code is distributed separately under the MIT License. See `NOTICE` for attribution. The third-party Cosmos Policy files in `libero/cosmos-2/` retain their original NVIDIA One-Way Noncommercial License (NSCLv1). See [the included license](libero/cosmos-2/LICENSE) and [original model card](libero/cosmos-2/UPSTREAM_README.md).