[torchlib] Implement type promotion by justinchuby · Pull Request #2010 · microsoft/onnxscript

justinchuby · 2025-01-15T00:29:03Z

Implement PyTorch type promotion for torchlib functions. Reference https://github.com/pytorch/pytorch/blob/bdd942efd76e74baa5dd0a262f7c843ddfe2e11b/torch/_prims_common/__init__.py#L1160

justinchuby · 2025-01-15T00:29:52Z

onnxscript/function_libs/torch_lib/ops/core.py

@@ -258,12 +263,13 @@ def aten_addmv(

 @torch_op("aten::addr", traceable=True)


traceonly etc.

codecov · 2025-01-15T00:32:00Z

❌ 59 Tests Failed:

Tests completed	Failed	Passed	Skipped
5743	59	5684	2733

View the top 3 failed tests by shortest run time

onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_0216_test_ceil

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_ceil'

The above exception was the direct cause of the following exception:
.nox\test\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_ceil' (e=No module named 'tests.onnx_backend_test_code.test_ceil') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_ceil.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_ceil.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
E   from onnxscript.onnx_opset import opset13
E   
E   @script()
E   def bck_test_ceil(x: FLOAT[3,4,5]) -> (FLOAT[3,4,5]):
E       y = opset13.Ceil(x)
E       return y

onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_0665_test_mod_mixed_sign_float16

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_mod_mixed_sign_float16'

The above exception was the direct cause of the following exception:
.nox\test\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_mod_mixed_sign_float16' (e=No module named 'tests.onnx_backend_test_code.test_mod_mixed_sign_float16') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_mod_mixed_sign_float16.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_mod_mixed_sign_float16.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 FLOAT16
E   from onnxscript.onnx_opset import opset13
E   
E   @script()
E   def bck_test_mod_mixed_sign_float16(x: FLOAT16[6], y: FLOAT16[6]) -> (FLOAT16[6]):
E       z = opset13.Mod(x, y, fmod=1)
E       return z

onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_1102_test_shape_example

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_shape_example'

The above exception was the direct cause of the following exception:
.nox\test\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_shape_example' (e=No module named 'tests.onnx_backend_test_code.test_shape_example') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_shape_example.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_shape_example.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 opset21
E   
E   @script()
E   def bck_test_shape_example(x: FLOAT[2,3]) -> (INT64[2]):
E       y = opset21.Shape(x)
E       return y

To view more test analytics, go to the Test Analytics Dashboard
📢 Thoughts on this report? Let us know!

onnxscript/function_libs/torch_lib/_type_promotion.py

@@ -0,0 +1,68 @@
+"""Type promotion functions for op implementations."""


onnxscript/function_libs/torch_lib/_type_promotion.py

@@ -0,0 +1,68 @@
+"""Type promotion functions for op implementations."""


onnxscript/function_libs/torch_lib/_type_promotion.py

@@ -0,0 +1,68 @@
+"""Type promotion functions for op implementations."""


onnxscript/function_libs/torch_lib/_type_promotion.py

@@ -0,0 +1,68 @@
+"""Type promotion functions for op implementations."""
+
+from typing import Sequence


titaiwangms · 2025-01-28T21:33:42Z

Should we merge a basic version before moving the whole thing to PyTorch so that we don't need to start from scratch?

justinchuby · 2025-01-28T22:03:27Z

Should we merge a basic version before moving the whole thing to PyTorch so that we don't need to start from scratch?

Sure! We can do that

xadupre · 2025-03-06T13:30:23Z

onnxscript/function_libs/torch_lib/_type_promotion.py

+    raise ValueError(f"Unexpected data types: {a}, {b}")
+
+
+def promote_types(op, values: Sequence[ir.Value]) -> Sequence[ir.Value]:


If it is used only for binary operators, I would add it to the function name just to make sure it is not called in any other operator.

justinchuby added 2 commits January 14, 2025 16:05

[torchlib] Implement type promotion

ecbb6d0

wip

b507bf5

justinchuby commented Jan 15, 2025

View reviewed changes

github-advanced-security bot found potential problems Jan 15, 2025

View reviewed changes

justinchuby added 4 commits January 15, 2025 12:48

, trace_only=True

1033ac5

format

eeeae08

TensorType

c6e300c

aten_addcdiv

fd915c1

justinchuby added 3 commits January 28, 2025 15:54

Merge branch 'main' into justinchu/type-promotion-reimpl

ff934c3

TensorType

5c45f70

Merge branch 'main' into justinchu/type-promotion-reimpl

db16b8b

xadupre reviewed Mar 6, 2025

View reviewed changes

justinchuby self-assigned this Jan 27, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[torchlib] Implement type promotion#2010

[torchlib] Implement type promotion#2010
justinchuby wants to merge 9 commits intomainfrom
justinchu/type-promotion-reimpl

justinchuby commented Jan 15, 2025

Uh oh!

justinchuby Jan 15, 2025

Uh oh!

codecov bot commented Jan 15, 2025 •

edited

Loading

Uh oh!

Check warning

Check warning

Check warning

Check warning

titaiwangms commented Jan 28, 2025

Uh oh!

justinchuby commented Jan 28, 2025

Uh oh!

xadupre Mar 6, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		@@ -258,12 +263,13 @@ def aten_addmv(

		@torch_op("aten::addr", traceable=True)

		@@ -0,0 +1,68 @@
		"""Type promotion functions for op implementations."""

		@@ -0,0 +1,68 @@
		"""Type promotion functions for op implementations."""

		from typing import Sequence

		raise ValueError(f"Unexpected data types: {a}, {b}")


		def promote_types(op, values: Sequence[ir.Value]) -> Sequence[ir.Value]:

Conversation

justinchuby commented Jan 15, 2025

Uh oh!

justinchuby Jan 15, 2025

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Jan 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

❌ 59 Tests Failed:

Uh oh!

Check warning

Check warning

Check warning

Check warning

titaiwangms commented Jan 28, 2025

Uh oh!

justinchuby commented Jan 28, 2025

Uh oh!

xadupre Mar 6, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codecov bot commented Jan 15, 2025 •

edited

Loading