README.md
| 1 | --- |
| 2 | tags: |
| 3 | - reinforcement-learning |
| 4 | - robotics |
| 5 | - stable-baselines3 |
| 6 | --- |
| 7 | # NoveltyAdapt policy checkpoints |
| 8 | |
| 9 | Public model checkpoints for [NoveltyAdapt](https://github.com/helenlu66/NoveltyAdapt), corresponding to the `iros2026` code branch. |
| 10 | |
| 11 | Paths in this model repository match paths relative to `learning/policies/` in the code repository. The top-level domains are `coffee_box`, `coffee_drawer`, `kitchen`, and `nut_assembly`; the original task, algorithm, training configuration, and seed subfolders are preserved. |
| 12 | |
| 13 | Only model checkpoint ZIP files are included: `final_model.zip`, `best_model.zip`, and `recent_model.zip`. Training/evaluation logs, replay buffers, and executor files are not included. |
| 14 | |
| 15 | ## Download |
| 16 | |
| 17 | From the code repository root: |
| 18 | |
| 19 | ```bash |
| 20 | pip install huggingface_hub |
| 21 | hf download helenlu94/NoveltyAdapt-policies --include "*.zip" --local-dir learning/policies |
| 22 | ``` |
| 23 | |
| 24 | See the code repository for environment dependencies and policy visualization instructions. |
| 25 | |