Skip to content

Arm backend: Change W2L "input type" for Ethos-U55 test #13839

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
merged 2 commits into from
Sep 1, 2025
Merged

Arm backend: Change W2L "input type" for Ethos-U55 test #13839

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
18c293f
Arm backend: Change W2L "input type" for Ethos-U55 test
Aug 26, 2025
9d76d4a
Merge branch 'main' into marlin-w2l
zingo Sep 1, 2025
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
25 changes: 16 additions & 9 deletions backends/arm/test/models/test_w2l_arm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,6 @@ class TestW2L(unittest.TestCase):
input_frames = 400
num_features = 1

w2l = models.Wav2Letter(num_features=num_features).eval()
model_example_inputs = get_test_inputs(batch_size, num_features, input_frames)

all_operators = [
Expand All @@ -45,11 +44,17 @@ class TestW2L(unittest.TestCase):
"executorch_exir_dialects_edge__ops_aten_relu_default",
]

@staticmethod
def create_model(input_type: str = "waveform"):
return models.Wav2Letter(
num_features=TestW2L.num_features, input_type=input_type
).eval()


@pytest.mark.slow # about 3min on std laptop
def test_w2l_tosa_FP():
pipeline = TosaPipelineFP[input_t](
TestW2L.w2l,
TestW2L.create_model(),
TestW2L.model_example_inputs,
aten_op=[],
exir_op=TestW2L.all_operators,
Expand All @@ -62,7 +67,7 @@ def test_w2l_tosa_FP():
@pytest.mark.flaky
def test_w2l_tosa_INT():
pipeline = TosaPipelineINT[input_t](
TestW2L.w2l,
TestW2L.create_model(),
TestW2L.model_example_inputs,
aten_op=[],
exir_op=TestW2L.all_operators,
Expand All @@ -74,12 +79,14 @@ def test_w2l_tosa_INT():
@pytest.mark.slow
@common.XfailIfNoCorstone300
@pytest.mark.xfail(
reason="MLETORCH-1009: Wav2Letter fails on U55 due to unsupported conditions",
strict=False,
reason="Wav2Letter fails on U55 due to insufficient memory",
strict=True,
)
def test_w2l_u55_INT():
pipeline = EthosU55PipelineINT[input_t](
TestW2L.w2l,
# Use "power_spectrum" variant because the default ("waveform") has a
# conv1d layer with an unsupported stride size.
TestW2L.create_model("power_spectrum"),
TestW2L.model_example_inputs,
aten_ops=[],
exir_ops=[],
Expand All @@ -94,7 +101,7 @@ def test_w2l_u55_INT():
@pytest.mark.skip(reason="Intermittent timeout issue: MLETORCH-856")
def test_w2l_u85_INT():
pipeline = EthosU85PipelineINT[input_t](
TestW2L.w2l,
TestW2L.create_model(),
TestW2L.model_example_inputs,
aten_ops=[],
exir_ops=[],
Expand All @@ -108,7 +115,7 @@ def test_w2l_u85_INT():
@pytest.mark.slow
def test_w2l_vgf_INT():
pipeline = VgfPipeline[input_t](
TestW2L.w2l,
TestW2L.create_model(),
TestW2L.model_example_inputs,
aten_op=[],
exir_op=TestW2L.all_operators,
Expand All @@ -121,7 +128,7 @@ def test_w2l_vgf_INT():
@common.SkipIfNoModelConverter
def test_w2l_vgf_FP():
pipeline = VgfPipeline[input_t](
TestW2L.w2l,
TestW2L.create_model(),
TestW2L.model_example_inputs,
aten_op=[],
exir_op=TestW2L.all_operators,
Expand Down
Loading