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 12, 17:15 UTC
[CI/Build] Skip tests that require libcudart in test_lmcache_integration.py (#29228)
Signed-off-by: Randall Smith <ransmith@amd.com> Co-authored-by: Randall Smith <ransmith@amd.com>
rasmith committed
Nov 22, 2025 at 04:59 UTC
a4fdf2405c737843d1e95e406959f3e2e6bcf899
1 file changed
+15
tests/v1/kv_connector/unit/test_lmcache_integration.py
+15
@@ -9,6 +9,12 @@
9
# Assumption vs. Correctness Tests:
10
# these unit tests do *not* test correctness of LMCache-side or vLLM-side logic
11
# it is to ensure that assumptions LMCache makes about vLLM's interface are stable
12
+
13
+import pytest
14
+
15
+from vllm.platforms import current_platform
16
+
17
+
18
def assumes(obj, attr, is_callable=False, is_instance_of=None):
19
import inspect
20
from dataclasses import is_dataclass
@@ -48,6 +54,9 @@ def assumes(obj, attr, is_callable=False, is_instance_of=None):
54
assert isinstance(attr_value, is_instance_of), assumption_msg
55
56
57
+@pytest.mark.skipif(
58
+ current_platform.is_rocm(), reason="Requires libcudart.so, not available on ROCm"
59
+)
60
def test_multimodal_interface():
61
# protect against interface changes
62
from vllm.multimodal.inputs import PlaceholderRange
@@ -72,6 +81,9 @@ def test_multimodal_interface():
81
assert token_ids.tolist() == [0, 0, 0, 0, 4, 4369, 4369, 4369, 4369, 9]
82
83
84
+@pytest.mark.skipif(
85
+ current_platform.is_rocm(), reason="Requires libcudart.so, not available on ROCm"
86
+)
87
def test_config_interface():
88
# protect against interface changes
89
from vllm.config import VllmConfig
@@ -146,6 +158,9 @@ def test_config_interface():
158
)
159
160
161
+@pytest.mark.skipif(
162
+ current_platform.is_rocm(), reason="Requires libcudart.so, not available on ROCm"
163
+)
164
def test_request_interface():
165
# protect against interface changes
166
from types import NoneType