master

commit 520eb0476a70dfd35f419437671e854419abb8b6
parent db85f91cb1bc12ec0e3633b3c6343b466a0ce1ed
Author: miksa234 <milutin@popovic.xyz>
Date:   Sun, 27 Apr 2025 20:28:29 +0100

revert last changes

Diffstat:
M
rl_arb/rl_arb/reinforce.py
 | 
5
+----
M
rl_arb/rl_arb/rlearn.py
 | 
20
+++++++++-----------

2 files changed, 10 insertions(+), 15 deletions(-)
diff --git a/rl_arb/rl_arb/reinforce.py b/rl_arb/rl_arb/reinforce.py
@@ -86,10 +86,7 @@ class Reinforce:
         test_rewards_hist = []
 
         for itr in range(self.args['num_reinforce']):
-            if itr % 2 == 0:
-                at_block = 10
-            else:
-                at_block = 11
+            at_block = 10
             self.mdp.current_block = at_block
             self.mcts.mdp.current_block = at_block
 
diff --git a/rl_arb/rl_arb/rlearn.py b/rl_arb/rl_arb/rlearn.py
@@ -151,12 +151,11 @@ class AgentRLearn():
             mean_val = np.mean(vs[idxs])
 
             if block not in self.baseline_tracker:
-                self.baseline_tracker[block] = mean_val
+                self.baseline_tracker[block] = list(vs[idxs])
             else:
-                self.baseline_tracker[block] = 1/2 * (mean_val + self.baseline_tracker[block])
-
-            baseline[idxs] = self.baseline_tracker[block]
+                self.baseline_tracker[block].append(vs[idxs])
 
+            baseline[idxs] = np.mean(self.baseline_tracker[block])
 
         return baseline*np.array(gamma_factors)
 
@@ -248,13 +247,12 @@ class AgentRLearn():
             with open("values.pickle", "wb") as f:
                 pickle.dump(self.values, f)
 
-            test_values = self.test_model(iteration)
-            self.test_values.append(test_values)
-            average_value = float(np.mean(test_values))
-            with open("test_values.pickle", "wb") as f:
-                pickle.dump(self.test_values, f)
-
-            send_telegram_message(f"Average profit {average_value}")
+            #test_values = self.test_model(iteration)
+            #self.test_values.append(test_values)
+            #average_value = float(np.mean(test_values))
+            #with open("test_values.pickle", "wb") as f:
+            #    pickle.dump(self.test_values, f)
+            #send_telegram_message(f"Average profit {average_value}")
 
         if self.args['telegram']:
             send_telegram_message("DONE!")