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, 07:15 UTC
[Misc] Change tests/compile to use VLLM_V1 by default (#19302)
Signed-off-by: rzou <zou3519@gmail.com>
Richard Zou committed
Jun 8, 2025 at 04:06 UTC
3d64d366e067a8121d923d7306fbcac0867411c6
2 files changed
+2
-15
tests/compile/conftest.py
deleted
-15
@@ -1,15 +0,0 @@
1
-# SPDX-License-Identifier: Apache-2.0
2
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
3
-import pytest
4
-
5
-
6
-# TEST V1: this should be removed. Right now V1 overrides
7
-# all the torch compile logic. We should re-enable this
8
-# as we add torch compile support back to V1.
9
-@pytest.fixture(scope="function", autouse=True)
10
-def use_v0_only(monkeypatch):
11
- """
12
- Since this module is V0 only, set VLLM_USE_V1=0 for
13
- all tests in the module.
14
- """
15
- monkeypatch.setenv('VLLM_USE_V1', '0')
tests/compile/piecewise/test_simple.py
+2
@@ -13,6 +13,7 @@ from vllm.compilation.counter import compilation_counter
13
from vllm.compilation.decorators import support_torch_compile
14
from vllm.config import (CompilationConfig, CompilationLevel, VllmConfig,
15
set_current_vllm_config)
16
+from vllm.envs import VLLM_USE_V1
17
from vllm.utils import direct_register_custom_op
18
19
global_counter = 0
@@ -76,6 +77,7 @@ class SillyModel(nn.Module):
77
78
79
def _test_simple_piecewise_compile(*, use_inductor):
80
+ assert VLLM_USE_V1
81
82
vllm_config = VllmConfig(compilation_config=CompilationConfig(
83
level=CompilationLevel.PIECEWISE,