smooth_l1_loss

[sushmanthreddy]

close #21165

[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! 🎉👨‍💻

[Daniel4078]

Hello, I suppose you need to implement the backend of your function for every framework available to allow your function to pass tests, not just paddle and torch.

[Daniel4078]

by the way, you should have a look at the new guide line of how we should write commit messages: https://www.conventionalcommits.org/en/v1.0.0/. Thanks.

[sushmanthreddy]

@Daniel4078 The problem here is smooth_l1_loss is and mixed function due to that only some backends have the implementation of smooth_l1_loss,oneis torch and paddle ,so that is the reason have implemented in this way.
link

[Daniel4078]

I would say that you should still try to implement this function compositionally in other backend by recreating its intended behavior using combinations of other functions in the target backend framework.

[sushmanthreddy]

i will work on it

[sushmanthreddy]

@Daniel4078 I have implemented the backend support manually in respective frameworks of jax,numpy and tensorflow...can u pls review that,it might have some mistakes over them,pls review it once and let me knew any changes are needed?

[Daniel4078]

Please update your branch with the main one as I found that your tests are failing due to an old problem: E ModuleNotFoundError: No module named 'dill'. Which has been patched out in recent commits on the ivy main branch.

[sushmanthreddy]

@Daniel4078 i have updated it and pushed the code..Now tests are passing as u mentiioned

screenshot of it is

[Daniel4078]

I think that it will be good for you slightly expand your docstring examples for your function like guided here(https://unify.ai/docs/ivy/overview/deep_dive/docstring_examples.html#docstring-examples). For example, some example requarding the use of reduction and out argument will be appreciated.

[sushmanthreddy]

@Daniel4078 i have changed the doc strings any new changes are needed??

[Daniel4078]

Found a failing test of your function, please have a look:

E   ivy.utils.exceptions.IvyBackendException: paddle: execute_with_gradients: paddle: nested_map: paddle: nested_map: (Fatal) There is no grad op for inputs:[0] or it'sstop_gradient=True.
E     [Hint: target_node should not be null.] (at /paddle/paddle/fluid/eager/general_grad.h:72)
E   
E   Falsifying example: test_smooth_l1_loss(
E       backend_fw='paddle',
E       on_device='cpu',
E       dtype_and_input=(['float64'], [array([-1., -1., -1.])]),
E       dtype_and_target=(['float64'], [array([-1., -1., -1.])]),
E       beta=0.5,
E       reduction='none',
E       test_flags=FunctionTestFlags(
E           ground_truth_backend='tensorflow',
E           num_positional_args=2,
E           with_out=False,
E           instance_method=False,
E           test_gradients=True,
E           test_compile=False,
E           as_variable=[False],
E           native_arrays=[False],
E           container=[False],
E       ),
E       fn_name='smooth_l1_loss',
E   )

[sushmanthreddy]

@Daniel4078 it has been failing for all of the loss function that is paddle bfloat16 issue or some backend issue.
If u check the for other loss functions ,which are implemented u can see that they will also fail for paddle only

[sushmanthreddy]

@Daniel4078 here is the screenshot of it.

[Daniel4078]

Maybe you can leave a comment in the relevent code saying this problem, then I can merge and let other ivy members to deal with this global problem.

[sushmanthreddy]

@Daniel4078 i have added in the doc string mentioning the error..can u pls review it ?

[Daniel4078]

Looks good for me!