Skip to content

Arm backend: Relax Qwen3 VL BF16 tolerance on Arm64 #20450

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zingo merged 1 commit into from
Jun 24, 2026
Merged

Arm backend: Relax Qwen3 VL BF16 tolerance on Arm64 #20450

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 9 additions & 2 deletions backends/arm/test/models/Qwen3_VL/test_qwen3_vl_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -246,15 +246,22 @@ def test_qwen3_vl_full_models_tosa_FP(test_case: Qwen3VLModelTestCase):
def test_qwen3_vl_full_models_tosa_FP_bf16(test_case: Qwen3VLModelTestCase):
model, inputs = test_case.model_cls.prepare_model_and_inputs()
model, inputs = _to_bfloat16_model_and_floating_inputs(model, inputs)
# Slightly higher atol for TOSA BF16 on aarch64 (MLETORCH-2048: numeric mismatch)
atol = (
0.4
if common.is_aarch64_host()
and test_case.model_cls is LowerableVisionModelWrapper
else 0.1
)
with torch.no_grad():
pipeline = TosaPipelineFP[input_t](
model,
inputs,
aten_op=[],
exir_op=[],
tosa_extensions=["bf16"],
atol=1e-1,
rtol=1e-1,
atol=atol,
rtol=0.1,
)
pipeline.run()

Expand Down
Loading