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, 12:15 UTC
[CI] Actually run tests/kv_transfer/test_disagg.py in CI (#17555)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
May 1, 2025 at 22:05 UTC
292fc59d61ba97b9a79b9649e693d0cac2dea15f
1 file changed
+2
-2
tests/kv_transfer/test_disagg.py
+2
-2
@@ -14,8 +14,8 @@ import torch
14
# Fixture to set up environment variables and teardown servers after tests
15
@pytest.fixture(scope="module", autouse=True)
16
def setup_servers():
17
- if torch.cuda.device_count() < 4:
18
- pytest.skip("Skipping test: fewer than 4 GPUs available")
17
+ if torch.cuda.device_count() < 2:
18
+ pytest.skip("Skipping test: fewer than 2 GPUs available")
19
20
# Set up environment variables
21
VLLM_HOST_IP = subprocess.check_output("hostname -I | awk '{print $1}'",