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 4, 18:45 UTC
[UT][Hardware] let torchrun example tests use the default backend (#39879)
Signed-off-by: zhenwei-intel <zhenwei.liu@intel.com>
liuzhenwei committed
Apr 23, 2026 at 16:22 UTC
4a79262e0f560c50a9831ee672a853e598047ee5
2 files changed
+4
-2
tests/distributed/test_torchrun_example.py
+2
-1
@@ -10,7 +10,8 @@ import torch.distributed as dist
10
from vllm import LLM, SamplingParams
11
from vllm.distributed.parallel_state import get_world_group
12
13
-dist.init_process_group(backend="gloo")
13
+# Let PyTorch choose the WORLD backend for the current device type.
14
+dist.init_process_group()
15
16
# Create prompts
17
prompts = [
tests/distributed/test_torchrun_example_moe.py
+2
-1
@@ -10,7 +10,8 @@ import torch.distributed as dist
10
from vllm import LLM, SamplingParams
11
from vllm.distributed.parallel_state import get_tp_group, get_world_group
12
13
-dist.init_process_group(backend="gloo")
13
+# Let PyTorch choose the WORLD backend for the current device type.
14
+dist.init_process_group()
15
16
# Create prompts
17
prompts = [