The tokenizer adds a special token '<|im_end|>' to solve the problem of non-stop generation when encountering <|im_end|>.

Using vllm to infer 'Llama3-ChatQA-1.5-70B', it will continue to be generated when encountering the special token '<|im_end|>', as shown in the figure below. This PR adds <|im_end|> to the tokenizer, and you need to add mapping to generation_config.json.

tokenizer.json

@@ -95,7 +95,7 @@
     },
     {
       "id": 128010,
-      "content": "<|reserved_special_token_5|>",
+      "content": "<|im_end|>",
       "single_word": false,
       "lstrip": false,
       "rstrip": false,