[Bugfix]: Fix messy code when using logprobs (#19209)

Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>

Author: chaunceyjiang

tests/test_utils.py

@@ -14,9 +14,12 @@
 import pytest
 import torch
 import zmq
+from transformers import AutoTokenizer
 from vllm_test_utils.monitor import monitor
 
 from vllm.config import ParallelConfig, VllmConfig, set_current_vllm_config
+from vllm.transformers_utils.detokenizer_utils import (
+    convert_ids_list_to_tokens)
 from vllm.utils import (CacheInfo, FlexibleArgumentParser, LRUCache,
                         MemorySnapshot, PlaceholderModule, StoreBoolean,
                         bind_kv_cache, common_broadcastable_dtype,
@@ -918,3 +921,14 @@ def test_split_host_port():
 def test_join_host_port():
     assert join_host_port("127.0.0.1", 5555) == "127.0.0.1:5555"
     assert join_host_port("::1", 5555) == "[::1]:5555"
+
+
+def test_convert_ids_list_to_tokens():
+    tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-1.5B-Instruct")
+    token_ids = tokenizer.encode("Hello, world!")
+    # token_ids = [9707, 11, 1879, 0]
+    assert tokenizer.convert_ids_to_tokens(token_ids) == [
+        'Hello', ',', 'Ġworld', '!'
+    ]
+    tokens = convert_ids_list_to_tokens(tokenizer, token_ids)
+    assert tokens == ['Hello', ',', ' world', '!']

tests/v1/engine/test_output_processor.py

@@ -35,7 +35,7 @@ def _ref_convert_id_to_token(
     Returns:
       String representation of input token id
     """
-    return tokenizer.convert_ids_to_tokens(token_id) or ""
+    return tokenizer.decode([token_id]) or ""
 
 
 @pytest.mark.parametrize(

vllm/transformers_utils/detokenizer_utils.py

@@ -78,6 +78,7 @@ def convert_prompt_ids_to_tokens(
 def convert_ids_list_to_tokens(
     tokenizer: AnyTokenizer,
     token_ids: list[int],
+    skip_special_tokens: bool = False,
 ) -> list[str]:
     """Detokenize the input ids individually.
 
@@ -89,8 +90,15 @@ def convert_ids_list_to_tokens(
       Python list of token string representations
     
     """
-    token_str_lst = tokenizer.convert_ids_to_tokens(token_ids)
-    _replace_none_with_empty(token_str_lst)  # type: ignore
+    token_str_lst = []
+    for token_id in token_ids:
+        token_str = tokenizer.decode(
+            [token_id],
+            skip_special_tokens=skip_special_tokens,
+        )
+        if token_str is None:
+            token_str = ""
+        token_str_lst.append(token_str)
     return token_str_lst