feat: add broadcast_shape to paddle frontend

[charlie572]

Related Issue

Close #22919

Checklist

• Did you add a function?
• Did you add the tests?
• Did you follow the steps we provided?

[charlie572]

The test doesn't pass, and I can't figure out why. It says this:

Expected :([1],)
Actual   :[1]

But when I run the function in the shell, I get the same result from both functions.

In [2]: from ivy.functional.frontends import paddle
In [3]: paddle.broadcast_shape((2, 1, 1), (1, 1, 2))
Out[3]: [2, 1, 2]
In [4]: paddle.broadcast_shape([2, 1, 1], [1, 1, 2])
Out[4]: [2, 1, 2]
In [5]: import paddle
In [6]: paddle.broadcast_shape((2, 1, 1), (1, 1, 2))
Out[6]: [2, 1, 2]
In [7]: paddle.broadcast_shape([2, 1, 1], [1, 1, 2])
Out[7]: [2, 1, 2]

Also, the function is supposed to allow you to input tuples or lists. I'm not sure how to test both of them in hypothesis. Do I have to sample from (list, tuple) with a strategy, and convert the shapes to that type?

[github-actions]

Thanks for contributing to Ivy! 😊👏
Here are some of the important points from our Contributing Guidelines 📝:
1. Feel free to ignore the run_tests (1), run_tests (2), … jobs, and only look at the display_test_results job. 👀 It contains the following two sections:
- Combined Test Results: This shows the results of all the ivy tests that ran on the PR. ✔️
- New Failures Introduced: This lists the tests that are passing on main, but fail on the PR Fork. Please try to make sure that there are no such tests. 💪
2. The lint / Check formatting / check-formatting tests check for the formatting of your code. 📜 If it fails, please check the exact error message in the logs and fix the same. ⚠️🔧
3. Finally, the test-docstrings / run-docstring-tests check for the changes made in docstrings of the functions. This may be skipped, as well. 📚
Happy coding! 🎉👨‍💻

[charlie572]

test_frontend_function wasn't checking list return values properly. I've tried to fix it.

[charlie572]

There are some tests failing because of types e.g. test_paddle_is_complex says. "Cannot convert to Paddle dtype. uint16 is not supported by Paddle." But I checked out previous commits, including the commit from just before I started this branch, and they fail for the same reason. Were these tests failing before, or have I set up my environment incorrectly? Also, I looked at the history of the main branch, and none of the commits in the last few weeks pass CI. So I'm confused about which tests I have to fix.

[MahmoudAshraf97]

can you please explain the changes you did in the testing files so @sherry30 can review them?

[github-actions]

PR Compliance Checks Passed!

[charlie572]

I've made some changes to test_frontend_function in function_testing.py. I've added shallow=False ton line 775. This is because if ret was a list, then that line would modify it in place, which we don't want. I've also moved a return statement on line 999. This is because the value was being modified before returned, which we don't want, so I moved it to before the part where it is modified.

[sherry30]

lets not move the test_values if statement, what we should do instead is tuplify ret as well so both ret and frontend_ret are both in tuples. This will make it easy to loop over the returns in the test function (Since some functions can also return tuples, so its easier to always return a tuple).
The other change looks good

Thanks

[charlie572]

lets not move the test_values if statement, what we should do instead is tuplify ret as well so both ret and frontend_ret are both in tuples. This will make it easy to loop over the returns in the test function (Since some functions can also return tuples, so its easier to always return a tuple). The other change looks good

Thanks

I've done it. I hope I've understood you correctly.

[charlie572]

Ignore those last two commits. I was looking at the wrong function.

[charlie572]

test_paddle_tensor_unsqueeze gives the error
AssertionError: returned dtype = bfloat16, ground-truth returned dtype = uint16

I noticed that paddle has a problem with setting the dtype as bfloat16. It doesn't work if you use the dtype argument to the Tensor constructor (it sets the dtype to int64 instead). But it does work if you use astype.

paddle.Tensor(np.arange(10), dtype="bfloat16")
Out[19]: 
Tensor(shape=[10], dtype=int64, place=Place(cpu), stop_gradient=True,
       [0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
paddle.Tensor(np.arange(10), dtype="bfloat16").astype("bfloat16")
Out[20]: 
Tensor(shape=[10], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
       [0., 1., 2., 3., 4., 5., 6., 7., 8., 9.])

Maybe this has something to do with the error. Should I try to fix it, or should I open an issue in the original paddle repository? I don't know how ivy converts to backend tensors, so I don't know how to fix it in ivy.

[github-actions]

Thank you for this PR, here is the CI results:

---

This pull request does not result in any additional test failures. Congratulations!

[MahmoudAshraf97]

Hello, are you still working on this?

[charlie572]

Yes. Is it ready to merge?

[MahmoudAshraf97]

closing as duplicate of #23574 which was already merged