apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] bad words support v1 scheduler and specifiy token ids (#3608)

Browse Source 
* support bad_words_token_ids

* docs

* fix test

* fix

* bad words support kvcache v1 and token ids

* fix
This commit is contained in:
Sunny-bot1
2025-08-26 11:14:51 +08:00
committed by GitHub
parent c43a4bec00
commit c68c3c4b8b
16 changed files with 420 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
fastdeploy/input/qwen_vl_processor.py
View File

@@ -212,6 +212,12 @@ class QwenVLProcessor(TextProcessor):
request["stop_token_ids"] = stop_seqs
request["stop_seqs_len"] = stop_seqs_len
bad_words = request.get("bad_words")
bad_words_token_ids = request.get("bad_words_token_ids")
if bad_words:
bad_words_token_ids = self.update_bad_words(bad_words, bad_words_token_ids)
request["bad_words_token_ids"] = bad_words_token_ids
if request.get("prompt"):
multimodal_data = request.get("multimodal_data")
if multimodal_data is None: