Skip to content

Commit 34f66b6

Browse files
authored
Skip ROCm MoE Quantization (#2191)
Add ROCm support skip for MoE quantization tests This commit introduces a conditional skip for the MoE quantization tests when running on ROCm, indicating that support for this feature is still under development. The change ensures that tests do not fail prematurely in environments lacking complete ROCm functionality. Test Plan: - Run tests in environments with and without ROCm support to verify the skip behavior.
1 parent 4ee2ee1 commit 34f66b6

File tree

1 file changed

+7
-0
lines changed

1 file changed

+7
-0
lines changed

test/quantization/test_moe_quant.py

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
import unittest
22

3+
import pytest
34
import torch
45
from parameterized import parameterized
56

@@ -32,6 +33,12 @@
3233
is_sm_at_least_90,
3334
)
3435

36+
if torch.version.hip is not None:
37+
pytest.skip(
38+
"ROCm support for MoE quantization is under development",
39+
allow_module_level=True,
40+
)
41+
3542

3643
class TestMoEQuantCompile(unittest.TestCase):
3744
DEFAULT_PARAMS = (512, 256, 8, 2) # hidden_dim, expert_dim, num_experts, top_k

0 commit comments

Comments
 (0)