| license: cc-by-4.0 | |
| tags: | |
| - robotics | |
| - imitation-learning | |
| - behavior-cloning | |
| - flow-matching | |
| # Flow Matching — Double Insert (sim) | |
| Behavior-cloning / flow-matching policy checkpoints for the simulated double-insert task, across multiple DAgger-style rounds. | |
| ## Contents | |
| This repository contains **36** PyTorch checkpoints (`.pth`), totaling roughly **164G**. | |
| Each subdirectory corresponds to one training run, named as: | |
| ``` | |
| <run-name>_seed<seed>_<timestamp>/epoch_<N>.pth | |
| ``` | |
| Some runs include multiple epochs. | |
| ## Usage | |
| ```python | |
| import torch | |
| state = torch.load("epoch_99.pth", map_location="cpu") | |
| ``` | |
| ## License | |
| Released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). | |