diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py
index 4288f53e6d..6a84f12b73 100644
--- a/unsloth/models/rl_replacements.py
+++ b/unsloth/models/rl_replacements.py
@@ -207,7 +207,7 @@ def grpo_trainer__get_per_token_logps(function_name, function):
     if  function_name != "_get_per_token_logps": return function
 
     def _get_per_token_logps(self, model, input_ids, attention_mask, logits_to_keep):
-        if os.environ.get('UNSLOTH_USE_NEW_MODEL', '0') == '1':
+        if os.environ.get('UNSLOTH_USE_NEW_MODEL', '0') == '0':
             return None # Unsloth efficient GRPO
         # Otherwise, calculate normally:
         if not hasattr(self, '_autocast_dtype'):
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index 7993a9a48e..9508d64889 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -137,8 +137,6 @@ def unsloth_base_fast_generate(
     try: kwargs["pixel_values"] = kwargs["pixel_values"].to(dtype)
     except: pass
 
-    print(kwargs.keys())
-
     # Mixed precision autocast
     if os.environ.get("UNSLOTH_FORCE_FLOAT32", "0") == "1": dtype = torch.float32
     with torch.inference_mode(), torch.autocast(device_type = "cuda", dtype = dtype):