apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-25 09:57:51 +08:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

[Bugfix] Align thinking_budget behavior with ERNIE reasoning flow (#6934)

Browse Source 
* [Bugfix] Align thinking_budget behavior with ERNIE reasoning flow

* [Docs] Fix thinking_budget markdown formatting

* [Test] Align ernie thinking budget test with process_request_dict
This commit is contained in:
jackyYang6
2026-03-23 14:15:55 +08:00
committed by GitHub
parent 7a78001be2
commit 634d23a38a
10 changed files with 663 additions and 285 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/input/v1/ernie4_5_vl_processor/ernie4_5_vl_processor.py
+9
View File
@@ -216,6 +216,11 @@ class Ernie4_5_VLProcessor(Ernie4_5Processor):
bad_words_token_ids = self.update_bad_words(bad_words, bad_words_token_ids)
request.sampling_params.bad_words_token_ids = bad_words_token_ids
logits_processors_args = self._prepare_think_stop_sentence(
getattr(request.sampling_params, "logits_processors_args", None) or {}, max_model_len
)
request.sampling_params.logits_processors_args = logits_processors_args
if request.prompt_token_ids:
messages = request.messages
if messages:
@@ -267,6 +272,10 @@ class Ernie4_5_VLProcessor(Ernie4_5Processor):
# 截断超过长度限制的prompt
if max_model_len is not None and len(request.prompt_token_ids) > max_model_len:
request.prompt_token_ids = request.prompt_token_ids[: max_model_len - 1]
logits_processors_args = self._update_thinking_prompt_state(
request.prompt_token_ids, getattr(request.sampling_params, "logits_processors_args", None) or {}
)
request.sampling_params.logits_processors_args = logits_processors_args
max_tokens = max_model_len - len(request.prompt_token_ids)
if getattr(request.sampling_params, "max_tokens", None) is None: