[CI] Remove VLLM_TEST_CLEAN_GPU_MEMORY to avoid environment variable pollution that causes unnecessary GPU detection, thereby slowing down test case execution. (#3446)

Signed-off-by: wangyu <410167048@qq.com>
Signed-off-by: [Your Name] <your.email@example.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

tests/entrypoints/openai_api/test_openai_audio_list_endpoints.py

@@ -9,7 +9,6 @@ Migrated from ``tests/e2e/online_serving/test_voxtral_tts.py`` to colocate with
 import os
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
-os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "0"
 
 import httpx
 import pytest

tests/entrypoints/openai_api/test_qwen3_tts_speaker_embedding.py

@@ -10,7 +10,6 @@ via pre-computed ECAPA-TDNN embeddings, bypassing ref_audio extraction.
 import os
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
-os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "0"
 
 import struct
 

tests/entrypoints/openai_api/test_qwen3_tts_websocket.py

@@ -16,7 +16,6 @@ from tests.helpers.runtime import OmniServerParams
 from tests.helpers.stage_config import get_deploy_config_path
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
-os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "0"
 
 MODEL = "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice"
 

tests/entrypoints/openai_api/test_serving_speech_batch.py

@@ -9,7 +9,6 @@ Validates bulk synthesis via the batch API with actual model inference.
 import os
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
-os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "0"
 
 import base64
 import struct