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 14, 05:30 UTC
[Misc] Remove unused `executor.apply_model` (#26215)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Oct 4, 2025 at 16:45 UTC
7c2e91c4e0414d076033d038b3c4acf236198cf3
1 file changed
+1
-12
vllm/executor/executor_base.py
+1
-12
@@ -7,8 +7,7 @@ from abc import ABC, abstractmethod
7
from functools import cached_property
8
from typing import Any, Awaitable, Callable, List, Optional, Set, Union
9
10
-import torch.nn as nn
11
-from typing_extensions import TypeVar, deprecated
10
+from typing_extensions import TypeVar
11
12
import vllm.platforms
13
from vllm.config import VllmConfig
@@ -127,16 +126,6 @@ class ExecutorBase(ABC):
126
self.collective_rpc("initialize_cache",
127
args=(num_gpu_blocks, num_cpu_blocks))
128
130
- @deprecated("`llm_engine.model_executor.apply_model` will no longer work "
131
- "in V1 Engine. Please replace with `llm_engine.apply_model` "
132
- "and set `VLLM_ALLOW_INSECURE_SERIALIZATION=1`.")
133
- def apply_model(self, func: Callable[[nn.Module], _R]) -> list[_R]:
134
- """
135
- Run a function directly on the model inside each worker,
136
- returning the result for each of them.
137
- """
138
- return self.collective_rpc("apply_model", args=(func, ))
139
-
129
@cached_property # Avoid unnecessary RPC calls
130
def supported_tasks(self) -> tuple[SupportedTask, ...]:
131
output = self.collective_rpc("get_supported_tasks")