[ROCm][Tests] Avoid pooling cleanup waits against live shared engines (#57864)
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> Co-authored-by: OpenAI Codex <noreply@openai.com>
Andreas Karatzas committed
Sep 21, 2026 at 00:21 UTC
76ee20c1ec5a9d6e094ecaecd46d71a5d591d05c
3 files changed
+8
tests/models/language/pooling/conftest.py
+2
@@ -33,6 +33,8 @@ def release_gpu_memory_between_tests():
33
gc.collect()
34
if torch.accelerator.is_available():
35
torch.accelerator.empty_cache()
36
+ if current_platform.is_rocm():
37
+ return
38
try:
39
wait_for_memory_to_settle()
40
except ValueError as e:
tests/models/language/pooling_mteb_test/conftest.py
+3
@@ -9,6 +9,7 @@ import torch
9
10
from tests.utils import wait_for_memory_to_settle
11
from vllm.logger import init_logger
12
+from vllm.platforms import current_platform
13
14
logger = init_logger(__name__)
15
@@ -26,6 +27,8 @@ def release_gpu_memory_between_tests():
27
gc.collect()
28
if torch.accelerator.is_available():
29
torch.accelerator.empty_cache()
30
+ if current_platform.is_rocm():
31
+ return
32
try:
33
wait_for_memory_to_settle()
34
except ValueError as e:
tests/models/multimodal/pooling/conftest.py
+3
@@ -9,6 +9,7 @@ import torch
9
10
from tests.utils import wait_for_memory_to_settle
11
from vllm.logger import init_logger
12
+from vllm.platforms import current_platform
13
14
logger = init_logger(__name__)
15
@@ -26,6 +27,8 @@ def release_gpu_memory_between_tests():
27
gc.collect()
28
if torch.accelerator.is_available():
29
torch.accelerator.empty_cache()
30
+ if current_platform.is_rocm():
31
+ return
32
try:
33
wait_for_memory_to_settle()
34
except ValueError as e: