feat(frontends): added the std method to paddle frontend

[samthakur587]

close #22151

[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! 🎉👨‍💻

[samthakur587]

hey @mohame54 can you please check this PR.I am not getting any error in local by testing it.

[samthakur587]

hey @hmahmood24 can you please review the PR. and let me know whether any changes required or not. So i can move to the next task. 🚀

[hmahmood24]

Hey @samthakur587, thanks a lot for the spot and this PR! I have suggested some minor changes. Let me know once you're done with them. Thanks! :)

[hmahmood24]

There is a std function in paddle_frontend now. We should use that here instead of ivy.std. The resulting code would look something like:

def std(self, axis=None, unbiased=True, keepdim=False, name=None):
        return paddle_frontend.std(self, axis=axis, unbiased=unbiased, keepdim=keepdim)

[samthakur587]

done , made the changes required i think now its ready to merge. 🚀

thank you

[hmahmood24]

@samthakur587 The tests are failing. I think because you're using a strategy from some other module to generate the inputs for your function which doesn't allow you to generate the valid dtypes that your function is valid for.

[samthakur587]

@samthakur587 The tests are failing. I think because you're using a strategy from some other module to generate the inputs for your function which doesn't allow you to generate the valid dtypes that your function is valid for.

hey @hmahmood24 i have tried to debug the test function and changed my composite _statistical_dtype_values the input dtype("valid") but still the numpy is taking the float16 dtype . at my last commit i checked that all test are passing but now i think some changed happend in numpy testing. thats why numpy not passing but others backend are passing.

same thing is happening in the stat.py std function that i am calling here as paddle frontend also in Var method and etc.

[github-actions]

PR Compliance Checks Passed!

[samthakur587]

hey @hmahmood24 i have tried to debug the ivy.std and i changed the composite function __statistical_dtype_values the helper dtype to ('valid') then all the backend are not passing bcz i have seen that the std at the paddle stat fronted include the unit8 and same as tensorflow and numpy so they returned the error as arrestion with dtype .

i am not found the actual bug that why it taking all float16 even i taken the valid dtype. can you please check.

[hmahmood24]

Hey @samthakur587, thanks for the update. Can you try removing the with_supported_dtypes decorator altogether and see if that works for you? Since this std method calls the paddle_frontend.std function which already has the with_supported_dtypes decorator applied to it here. Also, no need to change the _statistical_dtype_values since that is being used in numerous places so we shouldn't really be modifying it

[samthakur587]

Hi! as you said i have tried with removing the decorator @with_unsupported_dtypes({"2.5.1 and below": ("float16", "bfloat16")}, "paddle"). but its not worked. at the paddle stat.py the std function that is implemented is supporting the ("float64","float32","uint16") and uint16 not supported as mention in the paddle doc. and falling at all the backend with the dtype "bfloat16".

So i think we should keep the decorator on the tensor.std method.

[samthakur587]

I tried to check the ivy.std function. by running the tests at pytest -v ivy_tests/test_ivy/test_functional/test_core/test_statistical.py::test_std. it also falling in all the backend so i added the unsupported_dtype("bfloat","float16") after that all the backend passed but the numpy is not passing even i have also applied the same decortor to numpy backend.

@_scalar_output_to_0d_array
@with_unsupported_dtypes({"1.26.0 and below": ("bfloat16","float16")}, backend_version)
def std(
    x: np.ndarray,
    /,
    *,
    axis: Optional[Union[int, Sequence[int]]] = None,
    correction: Union[int, float] = 0.0,
    keepdims: bool = False,
    out: Optional[np.ndarray] = None,
) -> np.ndarray:
    axis = tuple(axis) if isinstance(axis, list) else axis
    return ivy.astype(np.std(x, axis=axis, ddof=correction, keepdims=keepdims, out=out),x.dtype,copy=False)


std.support_native_out = True

I don't konw but numpy still supporting the bfloat16 dtype. in the ivy.std also the numpy is not passing and returning the same error bfloat16 not supporting. here i have atteched the result.

[samthakur587]

Hi! @hmahmood24

Now its passing all the test with all the backend. and ready to merge 🚀

[hmahmood24]

Hey @samthakur587, looks good now. Can you just fix the lint issues with the PR so we can go ahead and merge it? Thanks!

[hmahmood24]

@samthakur587 Everything looks good to merge. Thanks! 🚀