Test fft normalization

[justinchuby]

The numbers seem closer-ish for fft_c2r, but don't really match.

[Copilot]

Copilot reviewed 1 out of 1 changed files in this pull request and generated no comments.

Comments suppressed due to low confidence (1)

onnxscript/function_libs/torch_lib/ops/fft.py:121

• Verify that op.Shape returns a scalar value when using start and end to extract the dimension. If op.Shape returns a tensor, explicitly extract its value to ensure proper normalization scaling.

scale = (op.CastLike(last_dim_size, self)) / op.CastLike(

To fix the issue, we should remove the unused import of ir from the onnxscript module. This will clean up the code and eliminate the unnecessary dependency. The change should be made on line 17, where the import statement is defined.

[codecov]

❌ 8 Tests Failed:

Tests completed	Failed	Passed	Skipped
14854	8	14846	2205

View the top 3 failed test(s) by shortest run time

onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_1271_test_tril

Stack Traces | 0.003s run time

onnxscript\backend\onnx_export_test.py:137: in extract_functions
    mod = importlib.import_module(import_name)
C:\hostedtoolcache\windows\Python\3.11.9\x64\Lib\importlib\__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
E   ModuleNotFoundError: No module named 'tests.onnx_backend_test_code.test_tril'

The above exception was the direct cause of the following exception:
.nox\test_onnx_weekly\Lib\site-packages\parameterized\parameterized.py:620: in standalone_func
    return func(*(a + p.args), **p.kwargs, **kw)
onnxscript\backend\onnx_export_test.py:271: in test_export2python_produces_correct_onnx_script_model
    functions = extract_functions(backend_test.name, code, self.test_folder)
onnxscript\backend\onnx_export_test.py:139: in extract_functions
    raise AssertionError(
E   AssertionError: Unable to import 'tests.onnx_backend_test_code.test_tril' (e=No module named 'tests.onnx_backend_test_code.test_tril') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_tril.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_tril.py', current folder: D:\a\onnxscript\onnxscript
E   ---- CONTENT --
E   import numpy
E   from onnx import TensorProto
E   from onnx.helper import make_tensor
E   from onnxscript import script, external_tensor
E   from onnxscript.values import Opset
E   from onnxscript.onnx_types import INT64
E   from onnxscript.onnx_opset import opset14
E   
E   @script()
E   def bck_test_tril(x: INT64[4,5]) -> (INT64[4,5]):
E       y = opset14.Trilu(x, upper=0)
E       return y

onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_0124_test_bitwise_not_4d

Stack Traces | 0.004s run time

onnxscript\backend\onnx_export_test.py:137: in extract_functions
    mod = importlib.import_module(import_name)
C:\hostedtoolcache\windows\Python\3.11.9\x64\Lib\importlib\__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
E   ModuleNotFoundError: No module named 'tests.onnx_backend_test_code.test_bitwise_not_4d'

The above exception was the direct cause of the following exception:
.nox\test_onnx_weekly\Lib\site-packages\parameterized\parameterized.py:620: in standalone_func
    return func(*(a + p.args), **p.kwargs, **kw)
onnxscript\backend\onnx_export_test.py:271: in test_export2python_produces_correct_onnx_script_model
    functions = extract_functions(backend_test.name, code, self.test_folder)
onnxscript\backend\onnx_export_test.py:139: in extract_functions
    raise AssertionError(
E   AssertionError: Unable to import 'tests.onnx_backend_test_code.test_bitwise_not_4d' (e=No module named 'tests.onnx_backend_test_code.test_bitwise_not_4d') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_bitwise_not_4d.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_bitwise_not_4d.py', current folder: D:\a\onnxscript\onnxscript
E   ---- CONTENT --
E   import numpy
E   from onnx import TensorProto
E   from onnx.helper import make_tensor
E   from onnxscript import script, external_tensor
E   from onnxscript.values import Opset
E   from onnxscript.onnx_types import UINT8
E   from onnxscript.onnx_opset import opset18
E   
E   @script()
E   def bck_test_bitwise_not_4d(x: UINT8[3,4,5,6]) -> (UINT8[3,4,5,6]):
E       bitwise_not = opset18.BitwiseNot(x)
E       return bitwise_not

onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_1153_test_slice_negative_axes

Stack Traces | 0.005s run time

onnxscript\backend\onnx_export_test.py:137: in extract_functions
    mod = importlib.import_module(import_name)
C:\hostedtoolcache\windows\Python\3.11.9\x64\Lib\importlib\__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
E   ModuleNotFoundError: No module named 'tests.onnx_backend_test_code.test_slice_negative_axes'

The above exception was the direct cause of the following exception:
.nox\test_onnx_weekly\Lib\site-packages\parameterized\parameterized.py:620: in standalone_func
    return func(*(a + p.args), **p.kwargs, **kw)
onnxscript\backend\onnx_export_test.py:271: in test_export2python_produces_correct_onnx_script_model
    functions = extract_functions(backend_test.name, code, self.test_folder)
onnxscript\backend\onnx_export_test.py:139: in extract_functions
    raise AssertionError(
E   AssertionError: Unable to import 'tests.onnx_backend_test_code.test_slice_negative_axes' (e=No module named 'tests.onnx_backend_test_code.test_slice_negative_axes') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_slice_negative_axes.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_slice_negative_axes.py', current folder: D:\a\onnxscript\onnxscript
E   ---- CONTENT --
E   import numpy
E   from onnx import TensorProto
E   from onnx.helper import make_tensor
E   from onnxscript import script, external_tensor
E   from onnxscript.values import Opset
E   from onnxscript.onnx_types import FLOAT, INT64
E   from onnxscript.onnx_opset import opset13
E   
E   @script()
E   def bck_test_slice_negative_axes(x: FLOAT[20,10,5], starts: INT64[3], ends: INT64[3], axes: INT64[3]) -> (FLOAT[20,10,1]):
E       y = opset13.Slice(x, starts, ends, axes)
E       return y

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[bmehta001]

Perhaps replace line 137 (op.Shape...) with op.CastLike(last_dim_size, self) and then remove scale? Would that yield the same/better results?

[justinchuby]

I thought last_dim_size was op.Shape(transformed, start=dimension, end=dimension + 1)? Let me try

[bmehta001]

Oh, nvm, sorry, you're completely right about op.Shape(transformed, start=dimension, end=dimension + 1) being different between line 122 and 137. But your code made me realize that without modifying anything else, line 137 perhaps should be directly replaced with last_dim_size just to save a call to op.Shape.

[bmehta001]

Perhaps simplify the normalization by passing in the dimension size you want directly to the normalization function. I think there is an underlying issue with my c2r implementation, though, because it should be able to support multiple axes. I can try to see if something like what is done in onnx/onnx#6016 will help

[justinchuby]

Perhaps simplify the normalization by passing in the dimension size you want directly to the normalization function. I think there is an underlying issue with my c2r implementation, though, because it should be able to support multiple axes. I can try to see if something like what is done in onnx/onnx#6016 will help

👍 looks like they recreated the conjugate part. It's like doubling the work that was meant to be saved but I guess that's the best option right now?