updated best results

llm_toolkit/llm_utils.py

@@ -183,11 +183,8 @@ def eval_model(
             decoded_output = tokenizer.batch_decode(
                 outputs, skip_special_tokens=True
             )  # Skip special tokens for clean output
-            debug = i == 0
-            decoded_output = [
-                extract_answer(output, debug=debug and j == 0)
-                for j, output in enumerate(decoded_output)
-            ]
+            if i == 0:
+                print("Batch output:", decoded_output)
             predictions.extend(decoded_output)
 
     return predictions