Update model card with checkpoint layout and release links
Browse files
README.md
CHANGED
|
@@ -1,31 +1,54 @@
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
library_name: pytorch
|
|
|
|
| 4 |
tags:
|
| 5 |
- robotics
|
| 6 |
- world-model
|
| 7 |
- planning
|
| 8 |
- pusht
|
| 9 |
- ogbench
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
arxiv: 2607.17973
|
| 11 |
---
|
| 12 |
|
| 13 |
-
# SAGE
|
| 14 |
|
| 15 |
-
Official checkpoints for [SAGE: Subgoal-Conditioned Action Generation for Latent World Model Planning](https://arxiv.org/abs/2607.17973).
|
| 16 |
|
| 17 |
-
|
| 18 |
|
| 19 |
-
|
| 20 |
-
| --- | --- | --- |
|
| 21 |
-
| `pusht_generator.pt` | PushT | Subgoal generator |
|
| 22 |
-
| `pusht_action_prior.pt` | PushT | Local-goal action prior |
|
| 23 |
-
| `pusht_far_action_prior.pt` | PushT | Far-goal action prior |
|
| 24 |
-
| `cube_generator.pt` | OGBench Cube | Subgoal generator |
|
| 25 |
-
| `cube_action_prior.pt` | OGBench Cube | Local-goal action prior |
|
| 26 |
-
| `cube_far_action_prior.pt` | OGBench Cube | Far-goal action prior |
|
| 27 |
|
| 28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
|
| 30 |
## Citation
|
| 31 |
|
|
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
library_name: pytorch
|
| 4 |
+
pipeline_tag: robotics
|
| 5 |
tags:
|
| 6 |
- robotics
|
| 7 |
- world-model
|
| 8 |
- planning
|
| 9 |
- pusht
|
| 10 |
- ogbench
|
| 11 |
+
- libero
|
| 12 |
+
- robotwin
|
| 13 |
+
datasets:
|
| 14 |
+
- CLTRAY/SAGE-data
|
| 15 |
arxiv: 2607.17973
|
| 16 |
---
|
| 17 |
|
| 18 |
+
# SAGE
|
| 19 |
|
| 20 |
+
Official checkpoints for **[SAGE: Subgoal-Conditioned Action Generation for Latent World Model Planning](https://arxiv.org/abs/2607.17973)**.
|
| 21 |
|
| 22 |
+
[Paper](https://arxiv.org/abs/2607.17973) | [Code](https://github.com/PKU-ML/SAGE) | [Datasets](https://huggingface.co/datasets/CLTRAY/SAGE-data)
|
| 23 |
|
| 24 |
+
SAGE combines latent subgoal generation and goal-conditioned action proposals with a frozen world model for planning.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
|
| 26 |
+
## Checkpoints
|
| 27 |
+
|
| 28 |
+
| Environment | Subgoal generator | Action prior | Far-goal action prior |
|
| 29 |
+
|:---|:---|:---|:---|
|
| 30 |
+
| PushT | `pusht_generator.pt` | `pusht_action_prior.pt` | `pusht_far_action_prior.pt` |
|
| 31 |
+
| OGBench Cube | `cube_generator.pt` | `cube_action_prior.pt` | `cube_far_action_prior.pt` |
|
| 32 |
+
|
| 33 |
+
| Environment | World model directory |
|
| 34 |
+
|:---|:---|
|
| 35 |
+
| LIBERO Scene2 | `libero_scene2/world_model/` |
|
| 36 |
+
| LIBERO Caddy | `libero_caddy/world_model/` |
|
| 37 |
+
| RoboTwin A2B | `robotwin_a2b/world_model/` |
|
| 38 |
+
|
| 39 |
+
Each world model directory contains `weights.pt` and `config.json`.
|
| 40 |
+
|
| 41 |
+
## Usage
|
| 42 |
+
|
| 43 |
+
Download the checkpoints with the Hugging Face Hub:
|
| 44 |
+
|
| 45 |
+
```python
|
| 46 |
+
from huggingface_hub import snapshot_download
|
| 47 |
+
|
| 48 |
+
snapshot_download(repo_id="CLTRAY/SAGE", local_dir="checkpoints")
|
| 49 |
+
```
|
| 50 |
+
|
| 51 |
+
See the [code repository](https://github.com/PKU-ML/SAGE) for installation and evaluation commands, and [native benchmark instructions](https://github.com/PKU-ML/SAGE/blob/main/NATIVE.md) for LIBERO and RoboTwin.
|
| 52 |
|
| 53 |
## Citation
|
| 54 |
|