From cd57f9c632b08a0b2f92dda785cf42cead30527e Mon Sep 17 00:00:00 2001
From: drbh <david.richard.holtz@gmail.com>
Date: Fri, 23 Feb 2024 14:53:18 +0000
Subject: [PATCH] fix: avoid duplicate bos token

---
 .../models/custom_modeling/flash_gemma_modeling.py              | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/server/text_generation_server/models/custom_modeling/flash_gemma_modeling.py b/server/text_generation_server/models/custom_modeling/flash_gemma_modeling.py
index 4a08bc2a..3161af9a 100644
--- a/server/text_generation_server/models/custom_modeling/flash_gemma_modeling.py
+++ b/server/text_generation_server/models/custom_modeling/flash_gemma_modeling.py
@@ -87,7 +87,7 @@ class GemmaTokenizerFast(PreTrainedTokenizerFast):
         bos_token="<bos>",
         eos_token="<eos>",
         pad_token="<pad>",
-        add_bos_token=True,
+        add_bos_token=False,
         add_eos_token=False,
         use_default_system_prompt=False,
         **kwargs,