[0.4.5] update DDPG

joyrl/__init__.py

@@ -5,13 +5,13 @@
 Email: [email protected]
 Date: 2023-01-01 16:20:49
 LastEditor: JiangJi
-LastEditTime: 2023-12-24 17:49:27
+LastEditTime: 2023-12-24 19:04:37
 Discription: 
 '''
 from joyrl import algos, framework, envs, utils
 from joyrl.run import run
 
-__version__ = "0.4.4.1"
+__version__ = "0.4.5"
 
 __all__ = [
     "algos",

joyrl/algos/DDPG/config.py

@@ -0,0 +1,32 @@
+#!/usr/bin/env python
+# coding=utf-8
+'''
+Author: JiangJi
+Email: [email protected]
+Date: 2023-12-15 13:16:24
+LastEditor: JiangJi
+LastEditTime: 2023-12-19 13:17:37
+Discription: 
+'''
+import numpy as np
+class AlgoConfig:
+    def __init__(self):
+        self.action_type = 'dpg' # action type, dpg: deterministic policy gradient
+        self.buffer_type = 'REPLAY_QUE' # replay buffer type
+        self.buffer_size = 100000  # replay buffer size
+        self.batch_size = 128  # batch size
+        self.gamma = 0.99  # discount factor
+        self.policy_loss_weight = 0.002  # policy loss weight
+        self.critic_lr = 1e-3  # learning rate of critic
+        self.actor_lr = 1e-4  # learning rate of actor
+        self.tau = 0.001  # soft update parameter
+        self.value_min = -np.inf  # clip min critic value
+        self.value_max = np.inf  # clip max critic value
+        self.actor_layers = [
+            {'layer_type': 'Linear', 'layer_size': [256], 'activation': 'ReLU'},
+            {'layer_type': 'Linear', 'layer_size': [256], 'activation': 'ReLU'},
+        ]
+        self.critic_layers = [
+            {'layer_type': 'Linear', 'layer_size': [256], 'activation': 'ReLU'},
+            {'layer_type': 'Linear', 'layer_size': [256], 'activation': 'ReLU'},
+        ]

joyrl/algos/DDPG/data_handler.py

@@ -0,0 +1,4 @@
+from joyrl.algos.base.data_handler import BaseDataHandler
+class DataHandler(BaseDataHandler):
+    def __init__(self, cfg):
+        super().__init__(cfg)