start with continue policy grad (wip)

MADEAPPS · Oct 17, 2023 · 3f922d2 · 3f922d2
1 parent e2cc2df
commit 3f922d2
Show file tree

Hide file tree

Showing 5 changed files with 497 additions and 24 deletions.
diff --git a/newton-4.00/applications/ndSandbox/demos/ndCartpoleDiscrete.cpp b/newton-4.00/applications/ndSandbox/demos/ndCartpoleDiscrete.cpp
@@ -22,7 +22,7 @@
 
 namespace ndCarpole_0
 {
-	//#define D_TRAIN_AGENT
+	#define D_TRAIN_AGENT
 
 	#define D_USE_VANILLA_POLICY_GRAD
 	//#define D_USE_PROXIMA_POLICY_GRAD
@@ -50,15 +50,15 @@ namespace ndCarpole_0
 		public:
 
 		#ifdef D_USE_VANILLA_POLICY_GRAD
-		class ndCartpoleAgent : public ndBrainAgentDiscretePolicyGrad<m_stateSize, m_actionsSize>
+		class ndCartpoleAgent : public ndBrainAgentPolicyGrad<m_stateSize, m_actionsSize>
 		#else
 		class ndCartpoleAgent : public ndBrainAgentDQN<m_stateSize, m_actionsSize>
 		#endif
 		{
 			public:
 			#ifdef D_USE_VANILLA_POLICY_GRAD
 			ndCartpoleAgent(ndSharedPtr<ndBrain>& actor)
-				:ndBrainAgentDiscretePolicyGrad<m_stateSize, m_actionsSize>(actor)
+				:ndBrainAgentPolicyGrad<m_stateSize, m_actionsSize>(actor)
 				,m_model(nullptr)
 			{
 			}