Update chess-gpt-eval-contrastive/main.py

chess-gpt-eval-contrastive/main.py

@@ -624,6 +624,8 @@ def play_games(
                 q_value_delta = curr_q_value - prev_q_value
                 material_bal = material_balance(board)
                 player_one.update_linear_probe_targets(curr_q_value, q_value_delta, material_bal)
+                player_one.train_linear_probes()
+                player_one.update_activations("reset")
             
             (
                 game_state,
@@ -677,8 +679,8 @@ def play_games(
             )
             games_saved += 1
 
-            if update_linear:
-                player_one.train_linear_probes()
+            #if update_linear:
+            #    player_one.train_linear_probes()
                 
             if update_activations:
                 if player_one_resignation or player_one_failed_to_find_legal_move: