Cap DreamWaQ adaptive learning rate

2026-07-22 19:58:19 +08:00
parent 944e91e28a
commit 509b3d635c
2 changed files with 9 additions and 1 deletions
--- a/motrix_rl/tests/test_rslrl_np_env_reward_flow.py
+++ b/motrix_rl/tests/test_rslrl_np_env_reward_flow.py
@@ -3,9 +3,14 @@ from types import SimpleNamespace
 import numpy as np
 import torch

+from motrix_rl.rslrl.torch.train.dreamwaq_ppo import DreamWaQPPO
 from motrix_rl.rslrl.torch.wrap_vec_env import RslrlNpEnvWrap


+def test_dreamwaq_adaptive_learning_rate_is_capped_at_initial_rate():
+    assert DreamWaQPPO.max_learning_rate == 1e-3
+
+
 def test_step_preserves_rewards_and_marks_only_truncations_as_timeouts():
    state = SimpleNamespace(
        obs=np.zeros((2, 3), dtype=np.float32),