Auto-sync: 2026-06-27 10:26:03 (part 2)
Browse files
scripts/eval_maniskill_policy_rollout.py
CHANGED
|
@@ -38,11 +38,12 @@ def main(argv: list[str] | None = None) -> int:
|
|
| 38 |
)
|
| 39 |
parser.add_argument(
|
| 40 |
"--selection-mode",
|
| 41 |
-
choices=("policy", "field", "lattice"),
|
| 42 |
default="policy",
|
| 43 |
help="'policy' executes the deterministic policy mean; 'field' scores model-generated "
|
| 44 |
"candidates with the learned interventional field; 'lattice' scores the current "
|
| 45 |
-
"state's CIL action lattice without reading rewards
|
|
|
|
| 46 |
)
|
| 47 |
parser.add_argument(
|
| 48 |
"--num-candidates",
|
|
|
|
| 38 |
)
|
| 39 |
parser.add_argument(
|
| 40 |
"--selection-mode",
|
| 41 |
+
choices=("policy", "field", "lattice", "retrieval_lattice"),
|
| 42 |
default="policy",
|
| 43 |
help="'policy' executes the deterministic policy mean; 'field' scores model-generated "
|
| 44 |
"candidates with the learned interventional field; 'lattice' scores the current "
|
| 45 |
+
"state's CIL action lattice without reading rewards; 'retrieval_lattice' scores "
|
| 46 |
+
"the nearest train-state lattice for the current state.",
|
| 47 |
)
|
| 48 |
parser.add_argument(
|
| 49 |
"--num-candidates",
|