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 16, 23:15 UTC
[mis] fix flaky test of test_cuda_device_count_stateless (#5546)
youkaichao committed
Jun 14, 2024 at 10:02 UTC
48f589e18b8b6758dbfb6bb23b2994430893b477
1 file changed
+2
-1
tests/distributed/test_utils.py
+2
-1
@@ -23,7 +23,8 @@ def test_cuda_device_count_stateless():
23
CUDA_VISIBLE_DEVICES is changed."""
24
25
actor = _CUDADeviceCountStatelessTestActor.options(num_gpus=2).remote()
26
- assert ray.get(actor.get_cuda_visible_devices.remote()) == "0,1"
26
+ assert sorted(ray.get(
27
+ actor.get_cuda_visible_devices.remote()).split(",")) == ["0", "1"]
28
assert ray.get(actor.get_count.remote()) == 2
29
ray.get(actor.set_cuda_visible_devices.remote("0"))
30
assert ray.get(actor.get_count.remote()) == 1