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, 22:35 UTC
[CI Bugfix] Fix CI OOM for `test_shared_storage_connector_hashes` (#21973)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Jul 30, 2025 at 23:45 UTC
055bd3978ededea015fb8f0cb6aa3cc48d84cde8
1 file changed
+3
-1
tests/v1/kv_connector/unit/test_shared_storage_connector.py
+3
-1
@@ -10,7 +10,7 @@ from vllm.assets.image import ImageAsset
10
from vllm.config import KVTransferConfig
11
from vllm.multimodal.utils import encode_image_base64
12
13
-MODEL_NAME = "Qwen/Qwen2.5-VL-3B-Instruct"
13
+MODEL_NAME = "RedHatAI/Qwen2.5-VL-3B-Instruct-quantized.w4a16"
14
15
SAMPLING_PARAMS = SamplingParams(temperature=0.0, top_k=1, max_tokens=128)
16
@@ -130,6 +130,8 @@ def test_shared_storage_connector_hashes(tmp_path):
130
model=MODEL_NAME,
131
max_model_len=8192,
132
max_num_seqs=1,
133
+ gpu_memory_utilization=0.4,
134
+ enforce_eager=True,
135
kv_transfer_config=kv_transfer_config,
136
limit_mm_per_prompt={"image": 2},
137
)