From 0a750da927ac977c3fd2bab9d039f77f7b6fb4db Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 21 Mar 2025 17:13:14 -0700
Subject: [PATCH] Update llama.py

---
 unsloth/models/llama.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 7557a96120..072d48780f 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -99,11 +99,13 @@ SDPA_HAS_GQA = "enable_gqa" in scaled_dot_product_attention.__doc__
 
 # Fix new HF's inference code
 def _fast_prepare_inputs_for_generation(self, input_ids, **kwargs,):
+    print("PREPARE", input_ids)
     if "past_key_values" in kwargs:
         input_ids = input_ids[:,[-1]]
         kwargs["attention_mask"] = kwargs["attention_mask"][:,[-1]]
     if "cache_position" in kwargs:
         kwargs["position_ids"] = kwargs["cache_position"]
+    print("PREPARE", input_ids)
     return { "input_ids" : input_ids, **kwargs, }
 pass