[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665)

Signed-off-by: Zach Zhu <zzqshu@126.com>
This commit is contained in:
Zach Zhu
2026-07-29 04:20:40 +00:00
committed by GitHub
parent f37f03db4a
commit 58f9659397
46 changed files with 370 additions and 248 deletions
+7 -1
View File
@@ -10,6 +10,7 @@ from typing import TYPE_CHECKING
import torch
from vllm.exceptions import VLLMValidationError
from vllm.logger import init_logger
from vllm.logits_process import LogitsProcessor as RequestLogitsProcessor
from vllm.sampling_params import SamplingParams
@@ -228,7 +229,12 @@ def validate_logits_processors_parameters(
tuple(logits_processors) if logits_processors is not None else None
)
for logits_procs in cached_load_custom_logitsprocs(logits_processors):
logits_procs.validate_params(sampling_params)
try:
logits_procs.validate_params(sampling_params)
except ValueError as e:
# Legacy custom logitsprocs may still raise ValueError from
# validate_params; convert for backward compatibility.
raise VLLMValidationError(str(e)) from e
class AdapterLogitsProcessor(LogitsProcessor):
+3 -1
View File
@@ -62,7 +62,9 @@ class LogitsProcessor(ABC):
def validate_params(cls, sampling_params: SamplingParams):
"""Validate sampling params for this logits processor.
Raise ValueError for invalid ones.
Raise ``VLLMValidationError`` (preferred) / ``ValueError`` (backward compatible)
for invalid params. Bare ``ValueError`` is converted to ``VLLMValidationError``
at the engine boundary so online serving returns HTTP 400.
"""
return None