Check weight shape dimensions in ConvTranspose shape inference msrc116345 by yuslepukhin · Pull Request #28524 · microsoft/onnxruntime

yuslepukhin · 2026-05-15T18:08:09Z

This pull request introduces comprehensive validation and error handling improvements for the ConvTranspose operator across CPU, CUDA, WebGPU, and XNNPACK backends, as well as in shape inference and unit tests. The main focus is to ensure that invalid input shapes (especially rank-0 or rank-1 tensors) are properly detected and reported, preventing undefined behavior and improving robustness. Additionally, error messages are clarified, and several helper functions now return Status for better error propagation.

Validation and Error Handling Improvements:

All ConvTranspose implementations (CPU, CUDA, WebGPU) now explicitly check that input X and filter W tensors have at least 3 dimensions, returning clear error messages if not. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R65-R79), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-d1bbcb0542b5acea587ac929cd6362cfd11172c522505c6db8b457a9d470c63dR273-R289), [[3]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-b615243d0702e9613bd815173108306495b0f690294001e606823b77322f6fafR22-L28))
The shape inference function for ConvTransposeWithDynamicPads now fails gracefully with descriptive errors if input or weight tensors have fewer than 2 dimensions. ([onnxruntime/core/graph/contrib_ops/contrib_defs.ccL62-R67](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-81f57d9adc2cce94f85a2949a895b7ff82efcc13d05e23ee6567661f0fecb7c0L62-R67))
Additional validation ensures that output_padding and dynamic pads have correct sizes, and that output_padding values are within ONNX-specified limits. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R138-R153), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R171-R187))

Refactoring for Robustness:

Helper functions such as ComputePadsAndOutputShape and ComputeTransposePadAndOutputShape now return Status, allowing errors to propagate and be handled appropriately rather than causing crashes or silent failures. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8L165-R234), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8L194-R262), [[3]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8L220-R282), [[4]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R291-R302))
All call sites (CPU, CUDA, WebGPU, XNNPACK) are updated to handle and propagate these errors using ORT_RETURN_IF_ERROR or ORT_THROW_IF_ERROR. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R171-R187), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-d1bbcb0542b5acea587ac929cd6362cfd11172c522505c6db8b457a9d470c63dL362-R379), [[3]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-b615243d0702e9613bd815173108306495b0f690294001e606823b77322f6fafL48-R60), [[4]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-6a2f8672090f25850b90b266aff3c7212552fc81b14bb7b539e9e5161c9fd526L494-R497))

Unit Test Enhancements:

New negative tests are added to verify that rank-0 and rank-1 weight tensors are properly rejected and produce the expected error messages, increasing test coverage and reliability. ([onnxruntime/test/contrib_ops/conv_transpose_with_dynamic_pads_test.ccR22-R56](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-cb5bfc51d0c8096922eb674d142f0e970d5becd140b47bdfd7729a06a818b598R22-R56))

Minor Code Quality Improvements:

Improved memory management in the CPU implementation by wrapping the allocated buffer in BufferUniquePtr immediately to prevent leaks if exceptions are thrown. ([onnxruntime/core/providers/cpu/nn/conv_transpose.ccR79-R89](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-0dcb5a9c8ba0c4e67940e9d77f77cb706bbf82d67bf6757967099b0a69c797b5R79-R89))
Minor includes and type safety improvements (e.g., use of SafeInt for overflow protection). ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R22), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R291-R302))

Summary of Most Important Changes:

1. Validation and Error Handling

All ConvTranspose implementations now check that input and filter tensors have at least 3 dimensions, returning clear errors if not. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R65-R79), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-d1bbcb0542b5acea587ac929cd6362cfd11172c522505c6db8b457a9d470c63dR273-R289), [[3]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-b615243d0702e9613bd815173108306495b0f690294001e606823b77322f6fafR22-L28))
Shape inference for ConvTransposeWithDynamicPads fails with descriptive errors for invalid input or weight tensor ranks. ([onnxruntime/core/graph/contrib_ops/contrib_defs.ccL62-R67](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-81f57d9adc2cce94f85a2949a895b7ff82efcc13d05e23ee6567661f0fecb7c0L62-R67))
Additional checks for output_padding and dynamic pads sizes and values, with ONNX spec compliance. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R138-R153), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R171-R187))

2. Error Propagation and Refactoring

Helper functions now return Status and propagate errors; all call sites updated to handle these errors. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8L165-R234), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8L194-R262), [[3]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8L220-R282), [[4]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R291-R302), [[5]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-d1bbcb0542b5acea587ac929cd6362cfd11172c522505c6db8b457a9d470c63dL362-R379), [[6]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-b615243d0702e9613bd815173108306495b0f690294001e606823b77322f6fafL48-R60), [[7]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-6a2f8672090f25850b90b266aff3c7212552fc81b14bb7b539e9e5161c9fd526L494-R497))

3. Unit Testing

Added tests to ensure invalid weight tensor ranks are rejected with proper error messages. ([onnxruntime/test/contrib_ops/conv_transpose_with_dynamic_pads_test.ccR22-R56](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-cb5bfc51d0c8096922eb674d142f0e970d5becd140b47bdfd7729a06a818b598R22-R56))

4. Code Quality

Improved buffer management and type safety in CPU backend. ([[1]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-0dcb5a9c8ba0c4e67940e9d77f77cb706bbf82d67bf6757967099b0a69c797b5R79-R89), [[2]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R22), [[3]](https://github.com/microsoft/onnxruntime/pull/28524/files#diff-72fa27d94d5d92dd1e78ff510ef9a84d1ad74426c19af9722cf6511f8d38a5a8R291-R302))

Copilot

Pull request overview

This PR targets improved robustness for the contrib op ConvTransposeWithDynamicPads by hardening its schema shape inference against invalid weight (W) ranks and adding regression tests for rank-0 and rank-1 weight inputs.

Changes:

Added a weight-shape rank guard in convTransposeWithDynamicPadsShapeInference before accessing W.dim(1).
Reused a cached weight_shape for computing the output channel dimension.
Added two unit tests covering rank-0 (scalar) and rank-1 weights expecting failure.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
`onnxruntime/core/graph/contrib_ops/contrib_defs.cc`	Adds a weight rank check in the `ConvTransposeWithDynamicPads` shape inference function and uses a cached `weight_shape` in channel inference.
`onnxruntime/test/contrib_ops/conv_transpose_with_dynamic_pads_test.cc`	Adds negative tests to ensure invalid `W` ranks are rejected.

Comments suppressed due to low confidence (1)

onnxruntime/test/contrib_ops/conv_transpose_with_dynamic_pads_test.cc:49

Same as the rank-0 case: please assert an expected failure message substring for this rank-1 test so it validates the intended weight-rank rejection path (rather than any downstream shape/attribute mismatch).

  test.AddOutput<float>("Y", {}, std::vector<float>{0.0f});
  test.Run(OpTester::ExpectResult::kExpectFailure, "", {kTensorrtExecutionProvider});
}

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 8 comments.

Comments suppressed due to low confidence (1)

onnxruntime/core/providers/cuda/nn/conv_transpose.cc:382

For dynamic_padding=true, pads are populated from the Pads input without validating that its element count matches 2*kernel_shape.size(). ComputePadsAndOutputShape writes to p_pads->at(dim) / at(rank+dim), so a too-short Pads tensor can trigger out_of_range/exception instead of a clean INVALID_ARGUMENT Status. Consider validating Pads rank/size up-front (similar to the CPU PrepareForCompute path) before calling ComputePadsAndOutputShape.

    ORT_RETURN_IF_ERROR(conv_transpose_attrs_.ComputePadsAndOutputShape(input_shape, num_output_channels, kernel_shape,
                                                                        strides, dilations, local_output_padding, N, &pads, &y_dims, channels_last));

    s_.y_dims = gsl::make_span(y_dims);
    s_.Y = context->Output(0, s_.y_dims);

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 6 comments.

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.

This pull request strengthens shape inference and error handling for the `MaxUnpool` and `ConvTranspose` ONNX operators. It adds stricter input validation to prevent silent failures or crashes when given invalid input shapes, and introduces new tests to ensure these cases are properly handled. **Shape inference validation improvements:** * Added a check in `maxUnpoolShapeInference` to ensure the indices input exists and has at least 2 dimensions, raising a shape inference error otherwise. * In `convTransposeShapeInference`, added explicit checks to ensure both the input tensor and the weight tensor have at least 2 dimensions, raising clear errors if not. * Updated logic in `convTransposeShapeInference` to consistently use the local `weight_shape` variable for shape calculations and checks, improving code clarity and correctness. [[1]](diffhunk://#diff-b9e98b88715e8a7e56c0dfa317f21a39582f919c87185f827c4a011ad9010639L1124-R1141) [[2]](diffhunk://#diff-b9e98b88715e8a7e56c0dfa317f21a39582f919c87185f827c4a011ad9010639L1194-R1206) **Test coverage enhancements:** * Added new unit tests to verify that shape inference raises appropriate errors for invalid input shapes, such as rank-0 or rank-1 indices/weights in `MaxUnpool` and `ConvTranspose`. **References** microsoft/onnxruntime#28524 microsoft/onnxruntime#28550 --------- Signed-off-by: Dmitri Smirnov <dmitrism@microsoft.com>

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 3 comments.

This pull request enhances the robustness of the `MaxUnpool` operator in ONNX Runtime by adding additional input validation and expanding test coverage for invalid input scenarios. The changes improve error handling for mismatched shapes, invalid dimensions, and other edge cases, ensuring the operator fails gracefully and predictably when given incorrect inputs. **Operator input validation improvements:** * Added runtime checks in `MaxUnpool::Compute` to ensure the `kernel_shape` rank matches the expected pooling dimensions, and that the indices tensor is present and correctly shaped. * Added validation to ensure that computed output dimensions are positive, with descriptive error messages if not. * Enforced that the `output_shape` tensor, if provided, must have the same number of elements as the rank of the input tensor. **Test coverage enhancements:** * Introduced multiple new tests in `unpool_op_test.cc` to cover invalid input cases, including mismatched indices shapes, rank-0 and rank-2 input tensors, negative indices, and incorrect `output_shape` element counts. These tests confirm that the operator fails with appropriate error messages in these scenarios. **References** onnx/onnx#7997 #28524

This pull request strengthens shape inference and error handling for the `MaxUnpool` and `ConvTranspose` ONNX operators. It adds stricter input validation to prevent silent failures or crashes when given invalid input shapes, and introduces new tests to ensure these cases are properly handled. **Shape inference validation improvements:** * Added a check in `maxUnpoolShapeInference` to ensure the indices input exists and has at least 2 dimensions, raising a shape inference error otherwise. * In `convTransposeShapeInference`, added explicit checks to ensure both the input tensor and the weight tensor have at least 2 dimensions, raising clear errors if not. * Updated logic in `convTransposeShapeInference` to consistently use the local `weight_shape` variable for shape calculations and checks, improving code clarity and correctness. [[1]](diffhunk://#diff-b9e98b88715e8a7e56c0dfa317f21a39582f919c87185f827c4a011ad9010639L1124-R1141) [[2]](diffhunk://#diff-b9e98b88715e8a7e56c0dfa317f21a39582f919c87185f827c4a011ad9010639L1194-R1206) **Test coverage enhancements:** * Added new unit tests to verify that shape inference raises appropriate errors for invalid input shapes, such as rank-0 or rank-1 indices/weights in `MaxUnpool` and `ConvTranspose`. **References** microsoft/onnxruntime#28524 microsoft/onnxruntime#28550 --------- Signed-off-by: Dmitri Smirnov <dmitrism@microsoft.com> Signed-off-by: Tommaso Adani <tommasoadani@microsoft.com>

Check weight shape dimensions in ConvTranspose shape inference

14a349b

yuslepukhin requested review from chilo-ms and Copilot May 15, 2026 18:08

Copilot started reviewing on behalf of yuslepukhin May 15, 2026 18:13 View session

Copilot AI reviewed May 15, 2026

View reviewed changes

Comment thread onnxruntime/core/graph/contrib_ops/contrib_defs.cc Outdated

Comment thread onnxruntime/test/contrib_ops/conv_transpose_with_dynamic_pads_test.cc

yuslepukhin added 2 commits May 15, 2026 11:52

Address specification and validation gaps

e95efd2

Make inference throw

1bf6435

edgchen1 reviewed May 18, 2026

View reviewed changes

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h Outdated

yuslepukhin added 2 commits May 18, 2026 11:24

Address SafeInt<> comments

4c9669e

Fix test issues

147d2df

yuslepukhin requested a review from edgchen1 May 18, 2026 18:51

yuslepukhin added 2 commits May 18, 2026 12:19

Fix test issues

1a3acdd

Fix test EP test issues

8374b8c

yuslepukhin requested a review from Copilot May 18, 2026 21:07

Copilot started reviewing on behalf of yuslepukhin May 18, 2026 21:08 View session

Copilot AI reviewed May 18, 2026

View reviewed changes

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose.cc Outdated

Address code review

8f07cad

yuslepukhin mentioned this pull request May 18, 2026

ConvTransposeWithDynamicPads msrc116345 #28548

Closed

yuslepukhin changed the title ~~Check weight shape dimensions in ConvTranspose shape inference~~ Check weight shape dimensions in ConvTranspose shape inference msrc116345 May 18, 2026

This was referenced May 18, 2026

Address shape inference shortcomings onnx/onnx#7997

Merged

Address MaxUnpool shortcomings msrc116345 #28550

Merged

Fix test EP test issues

0d1e93e

yuslepukhin requested a review from Copilot May 18, 2026 22:46

Copilot started reviewing on behalf of yuslepukhin May 18, 2026 22:48 View session

Copilot AI reviewed May 18, 2026

View reviewed changes

yuslepukhin added 2 commits May 18, 2026 16:50

Fix test EP test issues

fcd39a6

Addressing comments and merging PR#28548 into one.

46d77b6

yuslepukhin requested a review from Copilot May 19, 2026 00:09

Copilot AI reviewed May 19, 2026

View reviewed changes

Copilot started reviewing on behalf of yuslepukhin May 19, 2026 00:18 View session

yuslepukhin added 2 commits May 19, 2026 11:02

Addressing comments

ae014a3

Address CI failures.

09f204c

yuslepukhin requested a review from Copilot May 19, 2026 18:21

Copilot AI reviewed May 19, 2026

View reviewed changes

Comment thread onnxruntime/core/graph/contrib_ops/contrib_defs.cc

Comment thread onnxruntime/core/providers/webgpu/nn/conv_transpose.cc

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h

Copilot started reviewing on behalf of yuslepukhin May 19, 2026 18:40 View session

Address CI failures and review comments.

2a48fbc

yuslepukhin requested a review from Copilot May 19, 2026 20:58

Copilot AI reviewed May 19, 2026

View reviewed changes

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h

Comment thread onnxruntime/core/providers/webgpu/nn/conv_transpose.cc

Comment thread onnxruntime/test/providers/cuda/nhwc/conv_transpose_test.cc

Copilot started reviewing on behalf of yuslepukhin May 19, 2026 21:14 View session

Addressing comments

b9f3bbf

yuslepukhin requested review from gramalingam and justinchuby May 19, 2026 22:13

edgchen1 reviewed May 22, 2026

View reviewed changes

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose.cc Outdated

Comment thread onnxruntime/core/graph/contrib_ops/contrib_defs.cc

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h Outdated

yuslepukhin added 2 commits May 26, 2026 11:10

Adjust comment

174b9ad

Let SafeInt<> exception propagate

36cf9c4

edgchen1 previously approved these changes May 26, 2026

View reviewed changes

Comment thread onnxruntime/core/providers/cuda/nn/conv_transpose.cc Outdated

Comment thread onnxruntime/core/providers/cpu/nn/conv_transpose_attributes.h Outdated

Addressing comments

1062c9f

yuslepukhin dismissed edgchen1’s stale review via 1062c9f May 26, 2026 19:39

yuslepukhin enabled auto-merge (squash) May 26, 2026 20:36

edgchen1 approved these changes May 26, 2026

View reviewed changes

yuslepukhin merged commit 12916e9 into main May 26, 2026
88 checks passed

yuslepukhin deleted the yuslepukhin/shape_inference_msrc116345 branch May 26, 2026 23:00

Conversation

yuslepukhin commented May 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

yuslepukhin commented May 15, 2026 •

edited

Loading