Skip to content

Fix cross attention in MHA #2337

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.

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 4 commits into from
May 28, 2025
Merged

Fix cross attention in MHA #2337

merged 4 commits into from
May 28, 2025

Conversation

gramalingam
Copy link
Collaborator

Fix a seeming bug in handling of cross-attention in MHA (to be verified): In MHA fusion, we start with an input graph where attention is applied to 4D query/key/value, and it is transformed into a MHA op on 3D query/key/value.

In the case of cross-attention (with no rotary-embedding): the fusion seems to convert just query to 3D, and seems to leave key and value as 4D, which seems wrong.

This PR adds the necessary 4D=>3D conversion for key/value before MHA.

Note: This is a quick fix for the relevant case (that shows up). Other combinations may be worth checking out separately.

Signed-off-by: Ganesan Ramalingam <[email protected]>
Copy link

codecov bot commented May 23, 2025

❌ 3 Tests Failed:

Tests completed Failed Passed Skipped
15414 3 15411 2360
View the top 3 failed test(s) by shortest run time
onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_0129_test_bitwise_xor_i32_2d
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\3.12.10\x64\Lib\importlib\__init__.py:90: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
E   ModuleNotFoundError: No module named 'tests.onnx_backend_test_code.test_bitwise_xor_i32_2d'

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_bitwise_xor_i32_2d' (e=No module named 'tests.onnx_backend_test_code.test_bitwise_xor_i32_2d') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_bitwise_xor_i32_2d.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_bitwise_xor_i32_2d.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 INT32
E   from onnxscript.onnx_opset import opset18
E   
E   @script()
E   def bck_test_bitwise_xor_i32_2d(x: INT32[3,4], y: INT32[3,4]) -> (INT32[3,4]):
E       bitwisexor = opset18.BitwiseXor(x, y)
E       return bitwisexor
onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_0534_test_layer_normalization_4d_axis3
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\3.12.10\x64\Lib\importlib\__init__.py:90: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
E   ModuleNotFoundError: No module named 'tests.onnx_backend_test_code.test_layer_normalization_4d_axis3'

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_layer_normalization_4d_axis3' (e=No module named 'tests.onnx_backend_test_code.test_layer_normalization_4d_axis3') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_layer_normalization_4d_axis3.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_layer_normalization_4d_axis3.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 opset17
E   
E   @script()
E   def bck_test_layer_normalization_4d_axis3(X: FLOAT[2,3,4,5], W: FLOAT[5], B: FLOAT[5]) -> (FLOAT[2,3,4,5], FLOAT[2,3,4,1], FLOAT[2,3,4,1]):
E       Y, Mean, InvStdDev = opset17.LayerNormalization(X, W, B, axis=3)
E       return Y, Mean, InvStdDev
onnxscript.backend.onnx_export_test.TestOnnxBackEnd::test_export2python_produces_correct_onnx_script_model_0938_test_reshape_allowzero_reordered
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\3.12.10\x64\Lib\importlib\__init__.py:90: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
E   ModuleNotFoundError: No module named 'tests.onnx_backend_test_code.test_reshape_allowzero_reordered'

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_reshape_allowzero_reordered' (e=No module named 'tests.onnx_backend_test_code.test_reshape_allowzero_reordered') (file: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_reshape_allowzero_reordered.py', absolute path: 'D:\\a\\onnxscript\\onnxscript\\tests\\onnx_backend_test_code\\test_reshape_allowzero_reordered.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_reshape_allowzero_reordered(data: FLOAT[0,3,4], shape: INT64[3]) -> (FLOAT[3,4,0]):
E       reshaped = opset21.Reshape(data, shape, allowzero=1)
E       return reshaped

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

@gramalingam gramalingam enabled auto-merge (squash) May 28, 2025 00:43
@gramalingam gramalingam merged commit 5a8b9e6 into main May 28, 2025
22 of 29 checks passed
@gramalingam gramalingam deleted the rama/mha_cross branch May 28, 2025 01:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Development

Successfully merging this pull request may close these issues.

2 participants