Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,12 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Contents of datasets.
|
| 2 |
+
|
| 3 |
+
'param': All the parameters in the policy network as a flattened vector.
|
| 4 |
+
'traj': prior trajectories in first 60 steps, as 's_0, a_0, a_1, a_2, s_3, a_3, a_4, a_5'.
|
| 5 |
+
'task': the success three states
|
| 6 |
+
|
| 7 |
+
If you want to train with your dataset or task, you can privately design the trajectory dimensions and encode them to the same dimension (for example we used 128).
|
| 8 |
+
You can use our pretrained model with the same behavior dimensions to finetune on your dataset.
|
| 9 |
+
|
| 10 |
+
---
|
| 11 |
+
license: apache-2.0
|
| 12 |
+
---
|