xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-15 00:45:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

[PERF] Use faster way of decode in tokenizer: avoid useless list-to-list conversion (#20000)

Browse Source 
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@centml.ai>
This commit is contained in:
Vadim Gimpelson 2025-08-02 12:43:52 +04:00 committed by GitHub
parent 067c34a155
commit 58eee5f2e0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/transformers_utils/tokenizer.py
View File

@ -50,11 +50,12 @@ def decode_tokens(
`skip_special_tokens=None` means to use the backend's default
settings.
"""
decode_method = getattr(tokenizer, "_decode", tokenizer.decode)
if skip_special_tokens is not None:
return tokenizer.decode(token_ids,
return decode_method(token_ids,
skip_special_tokens=skip_special_tokens)
return tokenizer.decode(token_ids)
return decode_method(token_ids)
def encode_tokens(