Upload policy_config.py with huggingface_hub
Browse files- policy_config.py +27 -0
policy_config.py
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
exp_config = {
|
| 2 |
+
'type': 'ppo',
|
| 3 |
+
'on_policy': True,
|
| 4 |
+
'cuda': True,
|
| 5 |
+
'action_space': 'continuous',
|
| 6 |
+
'discount_factor': 0.99,
|
| 7 |
+
'gae_lambda': 0.95,
|
| 8 |
+
'epoch_per_collect': 10,
|
| 9 |
+
'batch_size': 64,
|
| 10 |
+
'learning_rate': 0.0003,
|
| 11 |
+
'lr_scheduler': None,
|
| 12 |
+
'weight_decay': 0,
|
| 13 |
+
'value_weight': 0.5,
|
| 14 |
+
'entropy_weight': 0.01,
|
| 15 |
+
'clip_ratio': 0.2,
|
| 16 |
+
'adv_norm': True,
|
| 17 |
+
'value_norm': 'baseline',
|
| 18 |
+
'ppo_param_init': True,
|
| 19 |
+
'grad_norm': 0.5,
|
| 20 |
+
'n_sample': 400,
|
| 21 |
+
'unroll_len': 1,
|
| 22 |
+
'deterministic_eval': True,
|
| 23 |
+
'model': {},
|
| 24 |
+
'cfg_type': 'PPOFPolicyDict',
|
| 25 |
+
'env_id': 'LunarLanderContinuous-v2',
|
| 26 |
+
'exp_name': 'LunarLanderContinuous-v2-PPO'
|
| 27 |
+
}
|