You are viewing a plain text version of this content. The canonical link for it is here.
Posted to discuss-archive@tvm.apache.org by "LudovicoYIN (via GitHub)" <gi...@apache.org> on 2026/01/16 06:08:59 UTC
[PR] [Relax][Onnx][Resize] Handle non-4D input tensors [tvm]
LudovicoYIN opened a new pull request, #18666:
URL: https://github.com/apache/tvm/pull/18666
### Motivation
The ONNX Resize operator supports resizing N-D tensors per the ONNX specification.
However, the current Relax ONNX frontend only supports 4D inputs and raises an
assertion error for valid non-4D models.
This PR extends the Relax ONNX Resize converter beyond the 4D-only restriction,
aligning behavior with the ONNX specification and ONNX Runtime for supported ranks.
### Changes
- Remove the 4D-only assertion in the Relax ONNX Resize converter
- Preserve the existing 4D resize path without behavior changes
- Support non-4D Resize by lowering to existing resize implementations for supported ranks
- Ensure Resize attributes are handled correctly for non-4D cases
### Testing
- Verified outputs against ONNX Runtime
- Added ONNX Resize tests covering non-4D input cases (`test_resize_nd_sizes`)
Fixes: [Bug] Resize N-D import failure: TVM only supports 4D resize2d, but ONNX Resize supports N-D tensors https://github.com/apache/tvm/issues/18608
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For additional commands, e-mail: discuss-archive-help@tvm.apache.org
Re: [PR] [Relax][Onnx][Resize] Handle non-4D input tensors [tvm]
Posted by "gemini-code-assist[bot] (via GitHub)" <gi...@apache.org>.
gemini-code-assist[bot] commented on PR #18666:
URL: https://github.com/apache/tvm/pull/18666#issuecomment-3758336837
## Summary of Changes
Hello @LudovicoYIN, I'm Gemini Code Assist[^1]! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request significantly enhances the Relax ONNX frontend's capability by extending the `Resize` operator to support N-dimensional input tensors, specifically 3D and 5D, in addition to the existing 4D support. This change brings the TVM implementation into full compliance with the ONNX specification, allowing for more flexible model imports and leveraging existing backend resize implementations for various tensor ranks.
### Highlights
* **Expanded ONNX Resize Support**: The Relax ONNX frontend now supports the `Resize` operator for 3D and 5D input tensors, in addition to the existing 4D support, aligning with the ONNX specification.
* **Generalized Attribute Handling**: The `mode` and `roi` attributes for the `Resize` operator have been generalized to correctly handle various tensor ranks, ensuring consistent behavior across different dimensions.
* **Conditional Resize Dispatch**: The implementation now dynamically dispatches to `topi.image.resize1d`, `relax.op.image.resize2d`, or `topi.image.resize3d` based on the input tensor's dimensionality.
* **New Test Cases**: New test cases (`test_resize_nd_sizes`) have been added to validate the correctness of the `Resize` operator for 1D, 2D, and 3D input tensors.
🧠**New Feature in Public Preview:** You can now enable **Memory** to help **Gemini Code Assist** learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. **Click [here](https://codeassist.google/code-review/login) to enable Memory in your admin console.**
<details>
<summary><b>Using Gemini Code Assist</b></summary>
<br>
The full guide for Gemini Code Assist can be found on our [documentation page](https://developers.google.com/gemini-code-assist/docs/review-github-code), here are some quick tips.
<b>Invoking Gemini</b>
You can request assistance from Gemini at any point by creating a comment using either `/gemini <command>` or `@gemini-code-assist <command>`. Below is a summary of the supported commands on the current page.
Feature | Command | Description
--- | --- | ---
Code Review | `/gemini review` | Performs a code review for the current pull request in its current state.
Pull Request Summary | `/gemini summary` | Provides a summary of the current pull request in its current state.
Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help | `/gemini help` | Displays a list of available commands.
<b>Customization</b>
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a `.gemini/` folder in the base of the repository. Detailed instructions can be found [here](https://developers.google.com/gemini-code-assist/docs/customize-gemini-behavior-github).
<b>Limitations & Feedback</b>
Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with :thumbsup: and :thumbsdown: on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up [here](https://google.qualtrics.com/jfe/form/SV_2cyuGuTWsEw84yG).
<b>You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the [Gemini Code Assist IDE Extension](https://cloud.google.com/products/gemini/code-assist).</b>
</details>
[^1]: Review the [Privacy Notices](https://policies.google.com/privacy), [Generative AI Prohibited Use Policy](https://policies.google.com/terms/generative-ai/use-policy), [Terms of Service](https://policies.google.com/terms), and learn how to configure Gemini Code Assist in GitHub [here](https://developers.google.com/gemini-code-assist/docs/customize-gemini-behavior-github). Gemini can make mistakes, so double check it and [use code with caution](https://support.google.com/legal/answer/13505487).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For additional commands, e-mail: discuss-archive-help@tvm.apache.org
Re: [PR] [Relax][Onnx][Resize] Handle non-4D input tensors [tvm]
Posted by "gemini-code-assist[bot] (via GitHub)" <gi...@apache.org>.
gemini-code-assist[bot] commented on code in PR #18666:
URL: https://github.com/apache/tvm/pull/18666#discussion_r2697110932
##########
python/tvm/relax/frontend/onnx/onnx_frontend.py:
##########
@@ -2287,17 +2292,45 @@ def _impl_v18(cls, bb, inputs, attr, params):
else:
assert f"Type {type(size)} for size is currently unsupported."
- return relax.op.image.resize2d(
+ if ndims == 3:
+ return bb.emit_te(
+ topi.image.resize1d,
+ x,
+ roi,
+ sizes,
+ "NCW",
+ topi_mode,
+ coord_mode,
+ rounding_method,
+ cubic_coeff_a,
+ exclude_outside,
+ extrapolation_value,
+ )
+ if ndims == 4:
+ return relax.op.image.resize2d(
+ x,
+ size=relax.ShapeExpr(sizes),
+ roi=roi,
+ layout="NCHW",
+ method=relax_mode,
+ coordinate_transformation_mode=coord_mode,
+ rounding_method=rounding_method,
+ cubic_alpha=cubic_coeff_a,
+ cubic_exclude=exclude_outside,
+ extrapolation_value=extrapolation_value,
+ )
+ return bb.emit_te(
+ topi.image.resize3d,
x,
- size=relax.ShapeExpr(sizes),
- roi=roi,
- layout="NCHW",
- method=mode,
- coordinate_transformation_mode=coord_mode,
- rounding_method=rounding_method,
- cubic_alpha=cubic_coeff_a,
- cubic_exclude=exclude_outside,
- extrapolation_value=extrapolation_value,
+ roi,
+ sizes,
+ "NCDHW",
+ topi_mode,
+ coord_mode,
+ rounding_method,
+ cubic_coeff_a,
+ exclude_outside,
+ extrapolation_value,
)
Review Comment:

For better readability and to make the logic more explicit, consider using an `if/elif/else` structure to handle the different `ndims` cases. This also makes the code more robust in case the assertion on `ndims` is changed in the future.
```python
if ndims == 3:
return bb.emit_te(
topi.image.resize1d,
x,
roi,
sizes,
"NCW",
topi_mode,
coord_mode,
rounding_method,
cubic_coeff_a,
exclude_outside,
extrapolation_value,
)
elif ndims == 4:
return relax.op.image.resize2d(
x,
size=relax.ShapeExpr(sizes),
roi=roi,
layout="NCHW",
method=relax_mode,
coordinate_transformation_mode=coord_mode,
rounding_method=rounding_method,
cubic_alpha=cubic_coeff_a,
cubic_exclude=exclude_outside,
extrapolation_value=extrapolation_value,
)
else: # ndims == 5
return bb.emit_te(
topi.image.resize3d,
x,
roi,
sizes,
"NCDHW",
topi_mode,
coord_mode,
rounding_method,
cubic_coeff_a,
exclude_outside,
extrapolation_value,
)
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For additional commands, e-mail: discuss-archive-help@tvm.apache.org
Re: [PR] [Relax][Onnx][Resize] Handle non-4D input tensors [tvm]
Posted by "tlopex (via GitHub)" <gi...@apache.org>.
tlopex merged PR #18666:
URL: https://github.com/apache/tvm/pull/18666
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For additional commands, e-mail: discuss-archive-help@tvm.apache.org
Re: [PR] [Relax][Onnx][Resize] Handle non-4D input tensors [tvm]
Posted by "LudovicoYIN (via GitHub)" <gi...@apache.org>.
LudovicoYIN commented on PR #18666:
URL: https://github.com/apache/tvm/pull/18666#issuecomment-3760749856
@tlopex Appreciate the review, thanks!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For additional commands, e-mail: discuss-archive-help@tvm.apache.org