A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 17:15 UTC
[Refactor] Remove unnecessary backend parameter in structured output interface (#15317)
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
Aaron Pham committed
Mar 26, 2025 at 13:51 UTC
733e7c9e95f5b066ac420b00701eef7ea164a79e
1 file changed
+5
-5
vllm/v1/structured_output/__init__.py
+5
-5
@@ -2,7 +2,7 @@
2
from __future__ import annotations
3
4
import multiprocessing
5
-from concurrent.futures import Future, ThreadPoolExecutor
5
+from concurrent.futures import ThreadPoolExecutor
6
from typing import TYPE_CHECKING, Optional
7
8
from vllm.config import VllmConfig
@@ -57,13 +57,13 @@ class StructuredOutputManager:
57
raise ValueError(
58
f"Unsupported structured output backend: {backend_name}")
59
60
- grammar: Future[StructuredOutputGrammar] = self.executor.submit(
61
- self._async_create_grammar, request, self.backend)
60
+ grammar = self.executor.submit(self._async_create_grammar, request)
61
request.structured_output_request.grammar = grammar # type: ignore[assignment]
62
63
def _async_create_grammar(
65
- self, request: Request,
66
- backend: StructuredOutputBackend) -> StructuredOutputGrammar:
64
+ self,
65
+ request: Request,
66
+ ) -> StructuredOutputGrammar:
67
key = request.structured_output_request.structured_output_key # type: ignore[union-attr]
68
69
# Note that the request was validated in the engine core client,