Download supervised_idms/cube_single/idm.json from mwilcoxson/selfimprovingv2a: direct link, hf CLI and curl.
- Browser
- Download file 625 Bytes
-
https://huggingface.co/datasets/mwilcoxson/selfimprovingv2a/resolve/main/supervised_idms/cube_single/idm.json
- Command line
-
hf download hf://datasets/mwilcoxson/selfimprovingv2a/supervised_idms/cube_single/idm.json
-
curl -L -o idm.json https://huggingface.co/datasets/mwilcoxson/selfimprovingv2a/resolve/main/supervised_idms/cube_single/idm.json
625 Bytes
| { | |
| "description": "flow IDM (QCX's flow IDM architecture) trained in vp_dagger with the DAgger Real VP of source_bundle, on the corpus released as vp_datasets/<env>; evaluates the released Real VP", | |
| "env": "cube_single", | |
| "frame_shape": [ | |
| 64, | |
| 64, | |
| 3 | |
| ], | |
| "frame_stack": 3, | |
| "horizon": 5, | |
| "idm": { | |
| "action_dim": 5, | |
| "compute_dtype": "bfloat16", | |
| "encoder_large": true, | |
| "hidden_dims": [ | |
| 1024, | |
| 1024, | |
| 1024, | |
| 1024 | |
| ], | |
| "minimal_action_ids": null | |
| }, | |
| "round": 9, | |
| "source_bundle": "/data/locomotion/scratch/mwil/vp_dagger_20261003/runs/vp_dagger/cube_single/dagger_mix18_vp600/seed0/m1/planner/round9" | |
| } | |