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 6, 09:30 UTC
[CI][BugFix][AMD] Don't set VLLM_ROCM_USE_AITER anymore in test_rocm_aiter_topk since its not necessary (#36996)
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
rasmith committed
Mar 19, 2026 at 18:12 UTC
98ff0429175b98169e1ebffd5ff32d0635bd39cc
1 file changed
-4
tests/kernels/moe/test_rocm_aiter_topk.py
-4
@@ -10,7 +10,6 @@
10
# and the platform is not ROCm.
11
12
import importlib.util
13
-import os
13
14
import pytest
15
import torch
@@ -20,9 +19,6 @@ from vllm.platforms import current_platform
19
if not current_platform.is_rocm():
20
pytest.skip("This test can only run on ROCm.", allow_module_level=True)
21
23
-# This environment variable must be set so ops will be registered.
24
-os.environ["VLLM_ROCM_USE_AITER"] = "1"
25
-
22
# this import statement is needed to ensure the ops are registered
23
import vllm.model_executor.layers.fused_moe.rocm_aiter_fused_moe # noqa: F401
24