feat: added unflatten frontend and backend support #27416

Kacper-W-Kozdon · 2023-11-30T20:25:34Z

feat: added unflatten Pytorch frontend and backend support

Currently added ivy.unflatten, unflatten for torch backend and unflatten in ivy/frontends/torch. Still requires unflatten (or the corresponding functions) in other backends.

Closes #22252

Checklist

Did you add a function?
Did you add the tests?
Did you run your tests and are your tests passing?
Did pre-commit not fail on any check?
Did you follow the steps we provided?

Socials

github-actions

PR Compliance Checks Passed!

Kacper-W-Kozdon · 2023-11-30T20:27:49Z

Reformatting Task Checklist

IMPORTANT NOTICE 🚨:

The Ivy Docs represent the ground truth for the task descriptions and this checklist should only be used as a supplementary item to aid with the review process.

LEGEND 🗺:

❌ : Check item is not completed.
✅ : Check item is ready for review.
🆘 : Stuck/Doubting implementation (PR author should add comments explaining why).
⏩ : Check is not applicable to function (skip).
🆗 : Check item is already implemented and does not require any edits.

CHECKS 📑:

- ❌: Remove all lambda and direct bindings for the backend functions in:
- ❌: Implement the following if they don't exist:
  1. - ❌: The ivy.Array instance method in ivy/data_classes/array/miscellaneous_operations.py.
  2. - ❌: The ivy.Array special method in ivy/data_classes/array/array.py.
  3. - ❌: The ivy.Array reverse special method in ivy/data_classes/array/array.py.
  4. - ❌: The ivy.Container static method in ivy/data_classes/container/miscellaneous_operations.py.
  5. - ❌: The ivy.Container instance method in ivy/data_classes/container/miscellaneous_operations.py.
  6. - ❌: The ivy.Container special method in ivy/data_classes/container/container.py.
  7. - ❌: The ivy.Container reverse special method in ivy/data_classes/container/container.py.
- ❌: Make sure that the aforementioned methods are added into the correct category-specific parent class, such as ivy.ArrayWithElementwise, ivy.ContainerWithManipulation etc.
- ❌: Correct all of the Function Arguments and the type hints for every function and its relevant methods, including those you did not implement yourself.
- ✅: Add the correct Docstrings to every function and its relevant methods, including those you did not implement yourself. The following should be added:
  1. - ❌: The function's Array API standard description in ivy/functional/miscellaneous_operations.py. If the function is not part of the Array API standard then a description of similar style should be added to the same file.
      The following modifications should be made to the description:
      - ✅: Remove type definitions in the Parameters and Returns sections.
      - ❌: Add out to the Parameters section if function accepts an out argument.
      - ✅: Replace out with ret in the Returns section.
  2. - ❌: Reference to docstring for ivy.function_name (5.a) for the function description and modified Parameters and Returns sections as described in the docs in:
      - ❌: ivy/array/miscellaneous_operations.py.
      - ❌: ivy/container/miscellaneous_operations.py (in the static and instance method versions).
      - ❌: ivy/array/array.py if the function has a special method ( like __function_name__ ).
      - ❌: ivy/array/array.py if the function has a reverse special method ( like __rfunction_name__ ).
      - ❌: ivy/container/container.py if the function has a special method ( like __function_name__ ).
      - ❌: ivy/container/container.py if the function has a reverse special method ( like __rfunction_name__ ).
- ✅: Add thorough Docstring Examples for every function and its relevant methods and ensure they pass the docstring tests.
  
  Functional Examples in ivy/functional/miscellaneous_operations.py.
  1. - ❌: Cover all possible variants for each of the arguments independently (not combinatorily).
  2. - ✅: Vary the values and input shapes considerably between examples.
  3. - ❌: Start out simple and get more complex with each example.
  4. - ❌: Show an example with:
      - ❌: out unused.
      - ❌: out used to update a new array y.
      - ❌: out used to inplace update the input array x (if x has the same dtype and shape as the return).
  5. - ❌: If broadcasting is relevant for the function, then show examples which highlight this.
  Nestable Function Examples in ivy/functional/miscellaneous_operations.py.
  Only if the function supports nestable operations.
  1. - ❌: Add an example that passes in an ivy.Container instance in place of one of the arguments.
  2. - ❌: Add an example passes in ivy.Container instances for multiple arguments.
  Container Static Method Examples in ivy/container/miscellaneous_operations.py.
  1. - ❌: The example from point (6.f) should be replicated, but added to the ivy.Container static method docstring in with ivy.<func_name> replaced with ivy.Container.static_<func_name> in the example.
  2. - ❌: The example from point (6.g) should be replicated, but added to the ivy.Container static method docstring, with ivy.<func_name> replaced with ivy.Container.static_<func_name> in the example.
  Array Instance Method Example in ivy/array/miscellaneous_operations.py.
  1. - ❌: Call this instance method of the ivy.Array class.
  Container Instance Method Example in ivy/container/miscellaneous_operations.py.
  1. - ❌: Call this instance method of the ivy.Container class.
  Array Operator Examples in ivy/array/array.py.
  1. - ❌: Call the operator on two ivy.Array instances.
  2. - ❌: Call the operator with an ivy.Array instance on the left and ivy.Container on the right.
  Array Reverse Operator Example in ivy/array/array.py.
  1. - ❌: Call the operator with a Number on the left and an ivy.Array instance on the right.
  Container Operator Examples in ivy/container/container.py.
  1. - ❌: Call the operator on two ivy.Container instances containing Number instances at the leaves.
  2. - ❌: Call the operator on two ivy.Container instances containing ivy.Array instances at the leaves.
  3. - ❌: Call the operator with an ivy.Container instance on the left and ivy.Array on the right.
  Container Reverse Operator Example in ivy/container/container.py.
  1. - ❌: Following example in the ivy.Container.__radd__ docstring, with the operator called with a Number on the left and an ivy.Container instance on the right.
  Tests
  1. - ❌: Docstring examples tests passing.
  2. - ✅: Lint checks passing.

Kacper-W-Kozdon · 2023-11-30T20:32:52Z

The frontend for unflatten is still missing most of the backends. Some backends use aliases. I'll implement that next.

joaozenobio · 2023-12-01T14:58:40Z

Hi @Kacper-W-Kozdon, thanks for starting a contribution! Ping me when you are ready to get a review or if you have any questions. 😁

…23279) Co-authored-by: NripeshN <nripesh14@gmail.com>

Co-authored-by: NripeshN <nripesh14@gmail.com>

… and node_split from Splitter

…ndom splitter for decision trees. check https://github.com/scikit-learn/scikit-learn/blob/a6603bcd8953e6bcc94081fa5c5d5741eb408927/sklearn/tree/_splitter.pyx#L860

…_feature_values

Co-authored-by: NripeshN <nripesh14@gmail.com>

Co-authored-by: Bhushan Srivastava <59949692+he11owthere@users.noreply.github.com>

…on call (Transpile-AI#27428) Co-authored-by: NripeshN <nripesh14@gmail.com>

Co-authored-by: NripeshN <nripesh14@gmail.com>

…dule derived from https://github.com/scikit-learn/scikit-learn/blob/c6654f919601ec54436f7a24a94e0874a763a8a1/sklearn/tree/_splitter.pyx#L289C20-L289C20

derived from https://github.com/scikit-learn/scikit-learn/blob/c6654f919601ec54436f7a24a94e0874a763a8a1/sklearn/tree/_splitter.pyx#L43

…splitter function

Kacper-W-Kozdon · 2024-01-02T23:21:47Z

Hi @Kacper-W-Kozdon ! Thanks for the modifications. From the pytest log, the error about the "dim" parameter is probably due to the generation of the values and as you observed a problem with the dtype_values_axis(), I'll ask you to try creating the these values separately using simpler helper functions, such as dtype_and_values(), get_axis(), etc.., and then after success we can prove there is a problem with the dtype_values_axis(). Feel free to pin me when you're done and Merry Christmas! 🎄🎁

There are still some problems with tests but splitting dtypes_values_axis() into dtype_and_values() + get_axis() seemingly helped with the dim issues.

UPDATE:
Running pytest on just the torch backend got a 100% pass- but it might be good to double-check it. If all's good, I can look into adding the rest of the backends in case they are still not implemented.
pytestdebug.log

Kacper-W-Kozdon · 2024-01-09T18:42:15Z

@joaozenobio the PR is ready for review.

Here dim and shape can be both positional and keyword. Ex: In: torch.unflatten(input=torch.zeros((2, 4, 2)), dim=1, sizes=(2, 2)).shape Out: torch.Size([2, 2, 2, 2])

joaozenobio

Hi @Kacper-W-Kozdon! Thanks for the updates 😄 I have tested your pytorch test and, if I did not got the wrong version of your code, the frontend test for torch did not pass. The error shown was:

x = tensor([[False, False]]), dim = 0, shape = (1, 2)
@handle_out_argument
def unflatten(
    x: torch.Tensor,
    /,
    dim: int = 0,
    shape: Tuple[int] = None,
    *,
    out: Optional[torch.Tensor] = None,
) -> torch.Tensor:
>       res = torch.unflatten(x, dim, shape)
E       RuntimeError: unflatten: Provided sizes [1, 2] don't multiply up to the size of dim 0 (1) in the input tensor

Could you please verify if the generation of this numbers are possible to be generated in your test?

Also, if you wish you can continue working with the other backends and also build a test function for testing the backend in this same PR 🚀

Kacper-W-Kozdon · 2024-01-16T21:55:35Z

Hi @Kacper-W-Kozdon! Thanks for the updates 😄 I have tested your pytorch test and, if I did not got the wrong version of your code, the frontend test for torch did not pass. The error shown was:
x = tensor([[False, False]]), dim = 0, shape = (1, 2)
@handle_out_argument
def unflatten(
    x: torch.Tensor,
    /,
    dim: int = 0,
    shape: Tuple[int] = None,
    *,
    out: Optional[torch.Tensor] = None,
) -> torch.Tensor:
>       res = torch.unflatten(x, dim, shape)
E       RuntimeError: unflatten: Provided sizes [1, 2] don't multiply up to the size of dim 0 (1) in the input tensor
Could you please verify if the generation of this numbers are possible to be generated in your test?

Also, if you wish you can continue working with the other backends and also build a test function for testing the backend in this same PR 🚀

No, you have the right one and I already know where to look for the cause. If the dimension to unflatten has size 1, there's nothing to factorise and the original shape was passed as the argument instead. This means that I need to add a condition for the dimension value equal to 1 and that should solve this problem. I will also look into other backends this week, it shouldn't take too much work. At the very least numpy and tensorflow have some functions similar to unflatten.

joaozenobio · 2024-01-16T22:39:09Z

Awesome! When you are done making the necessary modifications just pin me 😄

And thanks for the initiative! I am going to tag your PR to reference the other backends too 🚀

Kacper-W-Kozdon · 2024-01-20T13:44:12Z

Awesome! When you are done making the necessary modifications just pin me 😄

And thanks for the initiative! I am going to tag your PR to reference the other backends too 🚀

Test fixed (technically it should cover all needs as is, but after adding the missing backends I might move factorization function to some more fitting category if I find one and just for completeness- instead of full factorization into primes, get a list with just partial factors), all that was required was changing factors = [] to factors = [1], so now shape never gets an empty list.

Tensorflow backend for unflatten added- tensorflow doesn't have unflatten, so its behaviour was recreated with reshape.

Numpy seems to have unflatten natively, so I only have yet to check whichever backend was left (jax, I think?) from the ones considered in pytest. I'll have the PR ready by the end of the day.

Kacper-W-Kozdon · 2024-01-20T16:29:16Z

Awesome! When you are done making the necessary modifications just pin me 😄

And thanks for the initiative! I am going to tag your PR to reference the other backends too 🚀

Ready for review- all backends, pytorch frontend.
pytest log:
pytestdebug.log

joaozenobio · 2024-01-22T12:13:47Z

Nice work @Kacper-W-Kozdon ! All tests are passing and the superset behavior was well managed. Your PR is good to go 😄

joaozenobio · 2024-01-22T12:47:04Z

Hi @Kacper-W-Kozdon ! Your work with the frontend is great! With all the backends functions implemented, would you like to continue adding the features for this function implementing the Ivy Experimental API function and test and the data classes functions and tests, or would you like to finish our work here?

Kacper-W-Kozdon · 2024-01-22T12:52:10Z

Hi @Kacper-W-Kozdon ! Your work with the frontend is great! With all the backends functions implemented, would you like to continue adding the features for this function implementing the Ivy Experimental API function and test and the data classes functions and tests, or would you like to finish our work here?

Sure, I would happily look into that- but I'll have to get more familiar with the containers used by Ivy and how their methods are written. It shouldn't take too long but it'll be a touch slower than just the frontend implementation. Would it be possible to merge the current progress as is and open a new issue for the experimental API and classes?

joaozenobio · 2024-01-22T19:16:21Z

Sure thing! I'll merge your PR now and tag you in the new issue 😄

joaozenobio · 2024-01-25T21:54:20Z

Hi @Kacper-W-Kozdon ! Sorry for the delay, here is the new issue you should bond to your PR related to the next steps on the unflatten implementation. Thanks for the hard work!! 🚀

#28054

Draft for the ivy.unflatten | not finished

8a5020b

github-actions bot reviewed Nov 30, 2023

View reviewed changes

Kacper-W-Kozdon changed the title ~~feature: ivy.unflatten frontend (draft)~~ feat: ivy.unflatten frontend (draft) Nov 30, 2023

Kacper-W-Kozdon changed the title ~~feat: ivy.unflatten frontend (draft)~~ feat: ivy.unflatten frontend (not finished) Nov 30, 2023

ivy-leaves added PyTorch Frontend Developing the PyTorch Frontend, checklist triggered by commenting add_frontend_checklist Ivy Functional API labels Nov 30, 2023

ivy-seed assigned joaozenobio Nov 30, 2023

joaozenobio changed the title ~~feat: ivy.unflatten frontend (not finished)~~ feat: added ivy.unflatten Pytorch frontend and backend support Dec 1, 2023

joaozenobio changed the title ~~feat: added ivy.unflatten Pytorch frontend and backend support~~ feat: added unflatten Pytorch frontend and backend support Dec 1, 2023

AbdullahSabry and others added 18 commits December 4, 2023 19:04

feat: Added tridiagonal_solve in tensorflow frontend (Transpile-AI#…

48df5d3

…23279) Co-authored-by: NripeshN <nripesh14@gmail.com>

feat: added floor_mod in paddle frontend (Transpile-AI#26064)

ed14e0e

Co-authored-by: NripeshN <nripesh14@gmail.com>

feat: add base Splitter class for decision trees in sklearn front

f0e2366

🤖 Lint code

b892dd4

feat: add base structure for BestSplitter method which overrides init…

ffdf7f3

… and node_split from Splitter

🤖 Lint code

fa6ab8e

feat: add DensePartitioner required for both the best splitter and ra…

00a781f

…ndom splitter for decision trees. check https://github.com/scikit-learn/scikit-learn/blob/a6603bcd8953e6bcc94081fa5c5d5741eb408927/sklearn/tree/_splitter.pyx#L860

feat: add custom sort function for DensePartitions's sort_samples_and…

17325ed

…_feature_values

test commit (Transpile-AI#27421)

e66e1d5

feat: Added torchTensor.frac to torch frontend (Transpile-AI#27417)

cbeaa5e

Co-authored-by: NripeshN <nripesh14@gmail.com>

Add __getstate__ and __setstate__ methods to frontend torch.nn.Module

6b880f9

fix: Fixed typos (used codespell pre-commit hook) (Transpile-AI#27426)

a39a99a

Co-authored-by: Bhushan Srivastava <59949692+he11owthere@users.noreply.github.com>

fix: Fixed passing of arguments in get_referrers_recursive() functi…

f56f91d

…on call (Transpile-AI#27428) Co-authored-by: NripeshN <nripesh14@gmail.com>

feat: add erfc_ to torch frontend (Transpile-AI#27291)

392ec7c

Co-authored-by: NripeshN <nripesh14@gmail.com>

feat: add node_split_best function in sklearn frontend splitter submo…

7d77583

…dule derived from https://github.com/scikit-learn/scikit-learn/blob/c6654f919601ec54436f7a24a94e0874a763a8a1/sklearn/tree/_splitter.pyx#L289C20-L289C20

feat: define _init_split helper function used in node_split_best

888d200

derived from https://github.com/scikit-learn/scikit-learn/blob/c6654f919601ec54436f7a24a94e0874a763a8a1/sklearn/tree/_splitter.pyx#L43

feat: define SplitRecord class used for split structure in node best …

27a5f98

…splitter function

🤖 Lint code

827902b

Kacper-W-Kozdon added 2 commits January 3, 2024 00:44

unflatten_frontend tests fix

3702100

unflatten_frontend tests fix

4cd25ee

Kacper-W-Kozdon requested a review from joaozenobio January 3, 2024 00:40

Update manipulation.py

5100a1d

Here dim and shape can be both positional and keyword. Ex: In: torch.unflatten(input=torch.zeros((2, 4, 2)), dim=1, sizes=(2, 2)).shape Out: torch.Size([2, 2, 2, 2])

joaozenobio reviewed Jan 16, 2024

View reviewed changes

joaozenobio changed the title ~~feat: added unflatten Pytorch frontend and backend support~~ feat: added unflatten frontend and all backends support Jan 16, 2024

joaozenobio changed the title ~~feat: added unflatten frontend and all backends support~~ feat: added unflatten frontend and backend support Jan 16, 2024

Kacper-W-Kozdon and others added 5 commits January 19, 2024 13:13

Merge branch 'unifyai:main' into unflatten_frontend

92db59b

test- fix in factorization for dims

9ff2fd7

tensorflow backend experimental unflatten

8af48c6

tensorflow backend experimental unflatten

c0e8a00

tensorflow backend experimental unflatten

7a9a06d

Kacper-W-Kozdon added 3 commits January 20, 2024 16:21

numpy backend experimental unflatten

5fc20d5

paddle backend experimental unflatten

2e604a7

jax backend experimental unflatten

74c1be5

ivy-leaves added the Paddle Paddle Backend label Jan 20, 2024

Kacper-W-Kozdon requested a review from joaozenobio January 20, 2024 16:29

joaozenobio merged commit 12c965a into Transpile-AI:main Jan 22, 2024
357 of 413 checks passed

joaozenobio mentioned this pull request Jan 25, 2024

ivy.unflatten #28054

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: added unflatten frontend and backend support #27416

feat: added unflatten frontend and backend support #27416

Kacper-W-Kozdon commented Nov 30, 2023 •

edited by joaozenobio

github-actions bot left a comment •

edited

Kacper-W-Kozdon commented Nov 30, 2023 •

edited by joaozenobio

Kacper-W-Kozdon commented Nov 30, 2023

joaozenobio commented Dec 1, 2023 •

edited

Kacper-W-Kozdon commented Jan 2, 2024 •

edited

Kacper-W-Kozdon commented Jan 9, 2024

joaozenobio left a comment •

edited

Kacper-W-Kozdon commented Jan 16, 2024

joaozenobio commented Jan 16, 2024

Kacper-W-Kozdon commented Jan 20, 2024

Kacper-W-Kozdon commented Jan 20, 2024

joaozenobio commented Jan 22, 2024

joaozenobio commented Jan 22, 2024

Kacper-W-Kozdon commented Jan 22, 2024

joaozenobio commented Jan 22, 2024

joaozenobio commented Jan 25, 2024

feat: added unflatten frontend and backend support #27416

feat: added unflatten frontend and backend support #27416

Conversation

Kacper-W-Kozdon commented Nov 30, 2023 • edited by joaozenobio

feat: added unflatten Pytorch frontend and backend support

Checklist

Socials

github-actions bot left a comment • edited

Choose a reason for hiding this comment

Kacper-W-Kozdon commented Nov 30, 2023 • edited by joaozenobio

Reformatting Task Checklist

IMPORTANT NOTICE 🚨:

LEGEND 🗺:

CHECKS 📑:

Kacper-W-Kozdon commented Nov 30, 2023

joaozenobio commented Dec 1, 2023 • edited

Kacper-W-Kozdon commented Jan 2, 2024 • edited

Kacper-W-Kozdon commented Jan 9, 2024

joaozenobio left a comment • edited

Choose a reason for hiding this comment

Kacper-W-Kozdon commented Jan 16, 2024

joaozenobio commented Jan 16, 2024

Kacper-W-Kozdon commented Jan 20, 2024

Kacper-W-Kozdon commented Jan 20, 2024

joaozenobio commented Jan 22, 2024

joaozenobio commented Jan 22, 2024

Kacper-W-Kozdon commented Jan 22, 2024

joaozenobio commented Jan 22, 2024

joaozenobio commented Jan 25, 2024

Kacper-W-Kozdon commented Nov 30, 2023 •

edited by joaozenobio

github-actions bot left a comment •

edited

Kacper-W-Kozdon commented Nov 30, 2023 •

edited by joaozenobio

joaozenobio commented Dec 1, 2023 •

edited

Kacper-W-Kozdon commented Jan 2, 2024 •

edited

joaozenobio left a comment •

edited