| license: mit | |
| # LeWAM OGBench cube | |
| Goal-reaching checkpoint in the **LeWAM v1 format**. Files are stored at the repository root: | |
| ```text | |
| lewam_best.pt | |
| lewam_config.json | |
| ``` | |
| Download both files into `$STABLEWM_HOME/checkpoints/<run_name>/`, then evaluate with the v1 code: | |
| ```bash | |
| python scripts/eval_lewam.py --config-name cube policy=<run_name> | |
| ``` | |
| The checkpoint was trained with `lewam-jointflow@2d52888` and converted to the v1 schema. | |
| Every retained tensor is exactly equal to the original after renaming; strict model loading, | |
| attention masks, encoder output, supplied-goal action prediction and dynamics prediction passed. | |
| Only the obsolete, all-zero `null_goal` tensor was removed; the original goal-dropout setting was zero. | |
| No optimizer state is included. | |
| Original checkpoint and historical performance: [source repository](https://huggingface.co/datasets/LeWAM/lewam_ckpt_pusht_puzzle_scene_cube_drawer). | |
| Drawer uses goal offset 50 and budget 100; the other tasks use offset 25 and budget 50. | |
| SHA256 of `lewam_best.pt`: `73324009479d18256184d1ac3001254381fb6408aa213301a7c50f4923cf25e1`. | |