From 14a9768f0774b6abc63dce44daf08dc6364925c5 Mon Sep 17 00:00:00 2001
From: itazap <ita.zaporozhets@huggingface.co>
Date: Fri, 5 Dec 2025 10:52:50 +0100
Subject: [PATCH] qwen2 tok regression

---
 src/transformers/models/qwen2/tokenization_qwen2.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/src/transformers/models/qwen2/tokenization_qwen2.py b/src/transformers/models/qwen2/tokenization_qwen2.py
index 48312e3cadf8..86c4c458606c 100644
--- a/src/transformers/models/qwen2/tokenization_qwen2.py
+++ b/src/transformers/models/qwen2/tokenization_qwen2.py
@@ -63,7 +63,10 @@ def __init__(
             self._vocab = {
                 "<|endoftext|>": 0,
             }
-        self._merges = merges if merges is not None else generate_merges(self._vocab)
+        if merges is not None:
+            self._merges = [tuple(merge) if isinstance(merge, list) else merge for merge in merges]
+        else:
+            self._merges = generate_merges(self._vocab)
 
         self._tokenizer = Tokenizer(
             BPE(