use the original Qwen2.5-14B-Instruct tokenizer

source: https://huggingface.co/Qwen/Qwen2.5-14B-Instruct/tree/main

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
-size 11421896
+oid sha256:c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539
+size 7031645