apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-09-26 20:41:53 +08:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

delete default value reasoning_max_tokens (#4250)
Some checks failed
CE Compile Job / ce_job_pre_check (push) Has been cancelled
Details
CE Compile Job / print_ce_job_pre_check_outputs (push) Has been cancelled
Details
CE Compile Job / FD-Clone-Linux (push) Has been cancelled
Details
CE Compile Job / Show Code Archive Output (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8090 (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8689 (push) Has been cancelled
Details
CE Compile Job / CE_UPLOAD (push) Has been cancelled
Details

Browse Source 
* delete default value reasoning_max_tokens

* Adjust max_tokens and reasoning_max_tokens logic
This commit is contained in:
Yuanle Liu
2025-09-26 10:42:27 +08:00
committed by GitHub
parent 213f15ef55
commit dcf633c4d9
3 changed files with 4 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
fastdeploy/engine/engine.py
View File

@@ -210,9 +210,6 @@ class LLMEngine:
request.get("max_tokens"),
),
)
if request.get("reasoning_max_tokens") is None:
default_reasoning_max_tokens = max(int(request.get("max_tokens") * 0.8), 1)
request.set("reasoning_max_tokens", default_reasoning_max_tokens)
min_tokens = request.get("min_tokens")
if input_ids_len + min_tokens >= self.cfg.max_model_len:
error_msg = (

2
fastdeploy/engine/sampling_params.py
View File

@@ -159,8 +159,6 @@ class SamplingParams:
def __post_init__(self):
if self.seed is None:
self.seed = random.randint(0, 922337203685477580)
if self.max_tokens is not None and self.reasoning_max_tokens is None:
self.reasoning_max_tokens = max(int(self.max_tokens * 0.8), 1)
self._verify_args()
def _verify_args(self) -> None:

4
fastdeploy/input/ernie4_5_vl_processor/ernie4_5_vl_processor.py
View File

@@ -255,6 +255,10 @@ class Ernie4_5_VLProcessor(Ernie4_5Processor):
request["prompt_token_ids"] = request["prompt_token_ids"][: max_model_len - 1]
if request.get("max_tokens") is None:
request["max_tokens"] = max(1, max_model_len - len(request["prompt_token_ids"]))
else:
request["max_tokens"] = min(max_model_len - len(request["prompt_token_ids"]), request["max_tokens"])
if request.get("reasoning_max_tokens") is None:
request["reasoning_max_tokens"] = max(int(request["max_tokens"] * 0.8), 1)
data_processor_logger.info(f"Processed request {request}")
return request