From 5bf108ebdaab86688538f8b2a770ea33ef4341a7 Mon Sep 17 00:00:00 2001
From: Daniel Han-Chen <danielhanchen@gmail.com>
Date: Sun, 21 Jan 2024 18:15:55 +1100
Subject: [PATCH] Update llama.py

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

diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index f55d71c188..747ef0793a 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -306,6 +306,7 @@ def LlamaDecoderLayer_fast_forward(
     bsz, q_len, hd = hidden_states.size()
 
     if (not self.training and q_len == 1):
+        print(1)
         # Self Attention
         residual = hidden_states
         hidden_states = fast_rms_layernorm_inference(self.input_layernorm, hidden_states)