hotfix: skip test_mean_half_precision_underflow on amd ci (#9476)

The global size is very large (781250 gidx) and the emulated version takes more than 1
minute to execute the kernel.
This commit is contained in:
qazal
2025-03-17 16:47:48 +08:00
committed by GitHub
parent 824c5f41ac
commit e26caf4c3a

View File

@@ -820,6 +820,7 @@ class TestAutoCastType(unittest.TestCase):
np.testing.assert_allclose(t.grad.numpy(), [1, 0])
@unittest.skipIf(Device.DEFAULT == "PYTHON", "very slow")
@unittest.skipIf(CI and Device.DEFAULT == "AMD", "very slow")
@unittest.skipIf(Device.DEFAULT == "WEBGPU", "Binding size is larger than the maximum storage buffer binding size")
@unittest.skipUnless(is_dtype_supported(dtypes.half), "need half")
def test_mean_half_precision_underflow(self):