test: limit critic training to 5 steps for comparison purposes

Author: taoluo

tests/pipeline/rlvr_pipeline_meg_critic.py

@@ -347,7 +347,11 @@ def run(self):
         actor_infer_response_timer = _Timer(window_size=5)
         actor_train_timer = _Timer(window_size=5)
 
-        for global_step in range(self.pipeline_config.max_steps):
+        # Modified for testing: Run only 5 steps
+        test_max_steps = min(5, self.pipeline_config.max_steps)
+        logger.info(f"TEST MODE: Running only {test_max_steps} steps for critic comparison")
+        
+        for global_step in range(test_max_steps):
             if global_step <= self.state.step:
                 global_step += 1
                 continue