| license: other | |
| library_name: pytorch | |
| pipeline_tag: robotics | |
| tags: [long-wam, libero, cod] | |
| # Long-WAM LIBERO-COD | |
| Long-WAM policy weights for LIBERO. | |
| COD jointly denoises future-video latents and actions. | |
| ## Download | |
| ```bash | |
| hf download Efficient-Large-Model/Long-WAM-LIBERO-COD --local-dir ./weights/Long-WAM-LIBERO-COD | |
| ``` | |
| ## Inference | |
| Load `model.pt` with the matching Long-WAM runtime, using `config.yaml` and | |
| the supplied `dataset_stats.json` for preprocessing and normalization. | |
| P48 context uses 48 past control steps plus the current observation, encoded into four clean latent frames. | |
| Generate two future-video latents and a 32-step action chunk together using 4 joint denoising steps (sigma 1 to 0). | |
| Replan every 10 executed actions. Reset history at the start of each episode. | |
| The matching inference runtime and base-model VAE/text assets are required | |
| separately. This release contains weights and inference settings only; | |
| optimizer states and training artifacts are not included. | |