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, 12:15 UTC
[Misc] Remove deprecated files (#17447)
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Chauncey committed
Apr 30, 2025 at 16:52 UTC
1534d389af44a779d6b8b16f8574a50a2ab06f38
1 file changed
-35
vllm/model_executor/guided_decoding/reasoner/__init__.py
deleted
-35
@@ -1,35 +0,0 @@
1
-# SPDX-License-Identifier: Apache-2.0
2
-
3
-from __future__ import annotations
4
-
5
-from transformers import PreTrainedTokenizer
6
-
7
-from vllm.logger import init_logger
8
-from vllm.model_executor.guided_decoding.reasoner.deepseek_reasoner import ( # noqa: E501
9
- DeepSeekReasoner)
10
-from vllm.model_executor.guided_decoding.reasoner.reasoner import Reasoner
11
-
12
-logger = init_logger(__name__)
13
-
14
-
15
-def get_reasoner(tokenizer: PreTrainedTokenizer,
16
- reasoning_backend: str | None) -> Reasoner | None:
17
- if reasoning_backend is None:
18
- # No reasoning backend specified
19
- return None
20
- elif reasoning_backend == "deepseek_r1":
21
- return DeepSeekReasoner.from_tokenizer(tokenizer)
22
- elif reasoning_backend == "granite":
23
- logger.warning(
24
- "Granite reasoner not yet implemented for structured outputs")
25
- return None
26
- else:
27
- # Raise a warning for unknown reasoning backend and return None
28
- # We cannot raise an error here because some reasoning models
29
- # may not have a corresponding Reasoner class.
30
- logger.warning("Unknown reasoning backend %s for structured outputs ",
31
- reasoning_backend)
32
- return None
33
-
34
-
35
-__all__ = ["Reasoner", "get_reasoner"]