Skip to content

Commit e3eab48

Browse files
Arm backend: Remove Qwen3-VL BF16 patch embedding xfail (pytorch#21421)
PyTorch 2.13 resolves this AArch64 mismatch. Signed-off-by: Sebastian Larsson <sebastian.larsson@arm.com>
1 parent 1f7c6d0 commit e3eab48

1 file changed

Lines changed: 0 additions & 10 deletions

File tree

‎backends/arm/test/models/Qwen3_VL/test_qwen3_vl_layers.py‎

Lines changed: 0 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -510,16 +510,6 @@ def test_qwen3_vl_tosa_FP(test_case: Qwen3VLTestCase):
510510
@common.parametrize(
511511
"test_case",
512512
TOSA_BF16_TEST_CASES,
513-
xfails=(
514-
{
515-
"vision_patch_embed": (
516-
"MLETORCH-2048: Large bf16 patch embedding mismatch on aarch64",
517-
AssertionError,
518-
),
519-
}
520-
if common.is_aarch64_host()
521-
else None
522-
),
523513
)
524514
def test_qwen3_vl_tosa_FP_bf16(test_case: Qwen3VLTestCase):
525515
model, inputs = test_case.model_cls.prepare_model_and_inputs()

0 commit comments

Comments
 (0)