[Relax] Add native size operator #18667
Conversation
Summary of ChangesHello @guan404ming, I'm Gemini Code Assist1! 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 enhances the Relax framework by introducing a dedicated native operator, Highlights
🧠 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 to enable Memory in your admin console. Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization 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 Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 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. 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. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a native R.size(tensor) operator in Relax, which computes the total number of elements in a tensor. This is a useful addition, especially for ONNX model compatibility.
The changes are well-structured:
- A new
relax.sizeoperator is defined in C++ with struct info inference that correctly returns a scalarint64. - The operator is exposed to Python via FFI.
- A legalization rule is added to lower
relax.sizeto a sequence of existing ops (prod,shape_of,shape_to_tensor), which is a good design choice for reusing existing infrastructure. - The ONNX frontend is updated to use the new native operator.
- New tests are added for both static and dynamic shapes.
Overall, this is a solid contribution. I've added one suggestion to enhance the test coverage by including edge cases like scalar tensors and tensors with zero-sized dimensions.
| # Licensed to the Apache Software Foundation (ASF) under one | ||
| # or more contributor license agreements. See the NOTICE file | ||
| # distributed with this work for additional information | ||
| # regarding copyright ownership. The ASF licenses this file | ||
| # to you under the Apache License, Version 2.0 (the | ||
| # "License"); you may not use this file except in compliance | ||
| # with the License. You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, | ||
| # software distributed under the License is distributed on an | ||
| # "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY | ||
| # KIND, either express or implied. See the License for the | ||
| # specific language governing permissions and limitations | ||
| # under the License. | ||
|
|
||
| import numpy as np | ||
|
|
||
| import tvm | ||
| import tvm.testing | ||
| from tvm import relax | ||
| from tvm.script import relax as R | ||
|
|
||
|
|
||
| def test_op_size(): | ||
| @tvm.script.ir_module | ||
| class Module: | ||
| @R.function | ||
| def main(x: R.Tensor((2, 3), "float32")) -> R.Tensor((), "int64"): | ||
| return R.size(x) | ||
|
|
||
| x_np = np.random.rand(2, 3).astype("float32") | ||
| x = tvm.runtime.tensor(x_np) | ||
|
|
||
| target = tvm.target.Target("llvm") | ||
| ex = relax.build(Module, target) | ||
| vm = relax.VirtualMachine(ex, tvm.cpu()) | ||
|
|
||
| res = vm["main"](x) | ||
| assert res.numpy() == 6 | ||
|
|
||
|
|
||
| def test_op_size_dynamic(): | ||
| @tvm.script.ir_module | ||
| class Module: | ||
| @R.function | ||
| def main(x: R.Tensor(("m", "n"), "float32")) -> R.Tensor((), "int64"): | ||
| return R.size(x) | ||
|
|
||
| x_np = np.random.rand(4, 5).astype("float32") | ||
| x = tvm.runtime.tensor(x_np) | ||
|
|
||
| target = tvm.target.Target("llvm") | ||
| ex = relax.build(Module, target) | ||
| vm = relax.VirtualMachine(ex, tvm.cpu()) | ||
|
|
||
| res = vm["main"](x) | ||
| assert res.numpy() == 20 | ||
|
|
||
|
|
||
| if __name__ == "__main__": | ||
| tvm.testing.main() |
There was a problem hiding this comment.
The current tests cover static and dynamic shapes, which is great. To make the test suite more robust, consider adding tests for a few edge cases:
- Scalar input (0-dimensional tensor): The size should be 1.
- Tensor with a zero-sized dimension: The size should be 0.
Here are some examples of how you could write these tests:
def test_op_size_scalar():
@tvm.script.ir_module
class Module:
@R.function
def main(x: R.Tensor((), "float32")) -> R.Tensor((), "int64"):
return R.size(x)
x_np = np.array(1.0, dtype="float32")
x = tvm.runtime.tensor(x_np)
target = tvm.target.Target("llvm")
ex = relax.build(Module, target)
vm = relax.VirtualMachine(ex, tvm.cpu())
res = vm["main"](x)
assert res.numpy() == 1
def test_op_size_zero_dim():
@tvm.script.ir_module
class Module:
@R.function
def main(x: R.Tensor((2, 0, 3), "float32")) -> R.Tensor((), "int64"):
return R.size(x)
x_np = np.empty((2, 0, 3), dtype="float32")
x = tvm.runtime.tensor(x_np)
target = tvm.target.Target("llvm")
ex = relax.build(Module, target)
vm = relax.VirtualMachine(ex, tvm.cpu())
res = vm["main"](x)
assert res.numpy() == 0
1 participant
Why
ONNX models use the Size operator to get total element count of a tensor. Relax didn't have a native equivalent.
How