diff --git a/tests/models/quantization/test_gpt_oss.py b/tests/models/quantization/test_gpt_oss.py index fe9ddd2f6b..783f1773d2 100644 --- a/tests/models/quantization/test_gpt_oss.py +++ b/tests/models/quantization/test_gpt_oss.py @@ -22,7 +22,14 @@ import pytest from packaging import version from vllm.platforms import current_platform -from vllm.platforms.rocm import on_gfx950 + +if current_platform.is_rocm(): + from vllm.platforms.rocm import on_gfx950 +else: + + def on_gfx950() -> bool: + return False + MODEL_ACCURACIES = { # Full quantization: attention linears and MoE linears