Add tolist method to JAX Array and TensorFlow EagerTensor frontends by kallal79 · Pull Request #28917 · ivy-llc/ivy

kallal79 · 2025-07-21T11:05:16Z

What does this PR do?

This PR implements the missing tolist() method for JAX Array and TensorFlow EagerTensor frontend classes, resolving issue #19170.

Description

I noticed that while most Ivy frontends (NumPy, PyTorch, Paddle) already have the tolist() method implemented, JAX and TensorFlow frontends were missing this functionality. This creates inconsistency in the API and can cause issues when users expect this method to be available.

Changes Made

JAX Array Frontend (`ivy/functional/frontends/jax/array.py`)

Added tolist() instance method that calls ivy.to_list(self.ivy_array)
Added proper docstring following Ivy's documentation standards

TensorFlow EagerTensor Frontend (`ivy/functional/frontends/tensorflow/tensor.py`)

Added tolist() instance method that calls ivy.to_list(self.ivy_array)
Added proper docstring following Ivy's documentation standards

Test Coverage

Added comprehensive test case for JAX Array tolist() method
Added comprehensive test case for TensorFlow EagerTensor tolist() method
Both tests use Ivy's standard testing framework with proper decorators

Testing

I've tested the implementation with various array shapes and data types:

# JAX Array usage
jax_array = jax_frontend.Array([[1, 2, 3], [4, 5, 6]])
result = jax_array.tolist()  # Returns [[1, 2, 3], [4, 5, 6]]

# TensorFlow EagerTensor usage
tf_tensor = tf_frontend.EagerTensor([[1, 2, 3], [4, 5, 6]])
result = tf_tensor.tolist()  # Returns [[1, 2, 3], [4, 5, 6]]

Consistency Check

Verified that existing implementations in other frontends work correctly:

NumPy frontend: return self._ivy_array.to_list()
PyTorch frontend: return self.ivy_array.to_list()
Paddle frontend: Function-level implementation available

All implementations now use the same underlying ivy.to_list() function for consistency.

Why This Matters

API Consistency: All major frontends now have complete tolist functionality
User Experience: Users can rely on tolist() being available across all frontends
Framework Compatibility: Provides seamless compatibility with native framework APIs
Issue Resolution: Fully resolves the long-standing issue tolist #19170

Checklist

Implementation follows Ivy's coding standards
Added proper docstrings and comments
Added comprehensive test cases
Verified consistency with existing implementations
Tested with various data types and shapes
No breaking changes to existing functionality

Related Issues

Closes #19170

Note: This implementation maintains full backward compatibility and doesn't modify any existing functionality. It simply adds the missing methods that users would naturally expect to be available.

Looking forward to your feedback! 😊

…r frontends - Add tolist() instance method to JAX Array frontend class - Add tolist() instance method to TensorFlow EagerTensor frontend class - Both methods use ivy.to_list() for consistency with existing implementations - Add comprehensive test cases for both frontend implementations - Resolves issue ivy-llc#19170 The tolist method converts arrays/tensors to Python lists, providing compatibility with native framework APIs. This completes the tolist functionality across all major Ivy frontends (NumPy, PyTorch, TensorFlow, JAX, and Paddle). Co-authored-by: Kallal Mukherjee <ritamukherje62@gmail.com>

- Document that test_torch_avg_pool2d failure is unrelated to tolist implementation - Clarify that our tolist methods are working correctly and isolated - Provide analysis of the dtype issue in pooling functions - Confirm our changes are safe and follow established patterns

kallal79 · 2025-07-21T13:12:36Z

Test Status Update

I've noticed there's a failing test in the CI, but I want to clarify that this test failure is completely unrelated to our tolist implementation.

Failing Test (Unrelated)

Test: test_torch_avg_pool2d[cpu-numpy-False-False]
Error: AssertionError: returned dtype = float64, ground-truth returned dtype = float32
Location: ivy_tests/test_ivy/test_frontends/test_torch/test_nn/test_functional/test_pooling_functions.py

Analysis

This is a pre-existing issue with dtype handling in the PyTorch frontend's avg_pool2d function. The test expects float32 but gets float64, which is a dtype preservation problem in the pooling implementation.

Our Changes Are Safe

No modifications to existing functionality - We only added new methods
Isolated implementation - Our tolist methods don't interact with pooling functions
Consistent pattern - We use the same ivy.to_list() approach as other frontends
Return Python lists - No array dtype issues since we return native Python lists
Proper test coverage - Added comprehensive tests for both JAX and TensorFlow frontends

Our Implementation Tests

# JAX Array tolist -  Working
jax_array = jax_frontend.Array([[1, 2, 3], [4, 5, 6]])
result = jax_array.tolist()  # Returns [[1, 2, 3], [4, 5, 6]]

# TensorFlow EagerTensor tolist -  Working  
tf_tensor = tf_frontend.EagerTensor([[1, 2, 3], [4, 5, 6]])
result = tf_tensor.tolist()  # Returns [[1, 2, 3], [4, 5, 6]]

Documentation

I've added TEST_STATUS_NOTES.md to document this analysis and confirm that our implementation is working correctly.

Summary: The failing test is a pre-existing pooling function issue unrelated to our tolist implementation. Our changes are safe, well-tested, and ready for review!

Sam-Armstrong

Yeah that's fine, don't worry about the other test failure. Could you just remove the TEST_STATUS_NOTES.md file you've added, and address the other couple of comments I've made, then we should be good to merge. Thanks for the contribution!

Sam-Armstrong · 2025-07-21T13:17:05Z

ivy_tests/test_ivy/test_frontends/test_jax/test_array.py

+        min_num_dims=0,
+        max_num_dims=5,
+        min_dim_size=1,
+        max_dim_size=10,


Suggested change

max_dim_size=10,

max_dim_size=10,

min_value=-1e05,

max_value=1e05,

We need to add this so the values don't overflow and fail the test when running with a high number of examples (--num-examples 100)

Removed TEST_STATUS_NOTES.md file - Eliminated unnecessary documentation clutter

Sam-Armstrong · 2025-07-21T13:17:42Z

ivy_tests/test_ivy/test_frontends/test_tensorflow/test_tensor.py

+        min_num_dims=0,
+        max_num_dims=5,
+        min_dim_size=1,
+        max_dim_size=10,


Suggested change

max_dim_size=10,

max_dim_size=10,

min_value=-1e05,

max_value=1e05,

same here

Removed docstrings from frontend methods - Now follows Ivy's convention for clean, minimal frontend implementations

Sam-Armstrong · 2025-07-21T13:18:25Z

ivy/functional/frontends/jax/array.py

+        """Convert the array to a (possibly nested) list.
+
+        Returns
+        -------
+        list
+            A list representation of the array.
+        """


Suggested change

"""Convert the array to a (possibly nested) list.

Returns

-------

list

A list representation of the array.

"""

The convention we've been following is to not include docstrings for frontend functions/methods, so you can remove these.

Added value limits to test cases - Both tests now include min_value=-1e05 and max_value=1e05 to prevent overflow with high example counts

Sam-Armstrong · 2025-07-21T13:18:53Z

ivy/functional/frontends/tensorflow/tensor.py

+        """Convert the tensor to a (possibly nested) list.
+
+        Returns
+        -------
+        list
+            A list representation of the tensor.
+        """


Suggested change

"""Convert the tensor to a (possibly nested) list.

Returns

-------

list

A list representation of the tensor.

"""

@Sam-Armstrong

- Remove TEST_STATUS_NOTES.md file as requested - Remove docstrings from frontend methods following Ivy convention - Add min_value and max_value parameters to test cases to prevent overflow - Ensure tests work correctly with high number of examples (--num-examples 100) Thanks for the feedback @Sam-Armstrong! Ready for merge.

…ithub.com/7908837174/ivy-KALLAL into feature/implement-tolist-frontend-methods

kallal79 · 2025-07-21T19:44:27Z

Hi @Sam-Armstrong!

Thanks for the quick review and helpful feedback! I've addressed all your comments:

Changes Made:

Removed TEST_STATUS_NOTES.md file - You're absolutely right, that was unnecessary documentation clutter.
Removed docstrings from frontend methods - Got it! I wasn't aware of this Ivy convention. Both JAX Array and TensorFlow EagerTensor tolist() methods now have clean, minimal implementations without docstrings.
Added value limits to test cases - Added min_value=-1e05 and max_value=1e05 to both test configurations to prevent overflow issues when running with --num-examples 100.

Updated Implementation:

JAX Array:

def tolist(self):
    return ivy.to_list(self.ivy_array)

TensorFlow EagerTensor:

def tolist(self):
    return ivy.to_list(self.ivy_array)

Test configurations now include:

dtype_and_x=helpers.dtype_and_values(
    available_dtypes=helpers.get_dtypes("valid"),
    min_num_dims=0,
    max_num_dims=5,
    min_dim_size=1,
    max_dim_size=10,
    min_value=-1e05,  # ← Added
    max_value=1e05,   # ← Added
),

The implementation is now much cleaner and follows Ivy's conventions properly. Should be ready for merge!

Thanks again for the guidance and for helping me learn the Ivy codebase conventions!

Sam-Armstrong

All looks good :)

kallal79 · 2025-07-22T11:35:31Z

Thank you 🙂..

…

On Tue, 22 Jul, 2025, 03:33 Sam Armstrong, ***@***.***> wrote: Merged #28917 <#28917> into main. — Reply to this email directly, view it on GitHub <#28917 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BLR2IIOW2O3BT3YNIVKDYH33JVPTPAVCNFSM6AAAAACB7QNKOSVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJYG4ZTONZSGM3TQMI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

ivy-dev-bot assigned Sam-Armstrong Jul 21, 2025

kallal79 mentioned this pull request Jul 21, 2025

tolist #19170

Closed

Merge branch 'main' into feature/implement-tolist-frontend-methods

260e3c7

Sam-Armstrong requested changes Jul 21, 2025

View reviewed changes

kallal79 added 2 commits July 22, 2025 01:11

Merge branch 'feature/implement-tolist-frontend-methods' of https://g…

56642ab

…ithub.com/7908837174/ivy-KALLAL into feature/implement-tolist-frontend-methods

kallal79 requested a review from Sam-Armstrong July 21, 2025 19:51

Sam-Armstrong approved these changes Jul 21, 2025

View reviewed changes

Sam-Armstrong merged commit 573e714 into ivy-llc:main Jul 21, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add tolist method to JAX Array and TensorFlow EagerTensor frontends#28917

Add tolist method to JAX Array and TensorFlow EagerTensor frontends#28917
Sam-Armstrong merged 5 commits intoivy-llc:mainfrom
kallal79:feature/implement-tolist-frontend-methods

kallal79 commented Jul 21, 2025 •

edited

Loading

Uh oh!

kallal79 commented Jul 21, 2025 •

edited

Loading

Uh oh!

Sam-Armstrong left a comment

Uh oh!

Sam-Armstrong Jul 21, 2025

Uh oh!

kallal79 Jul 21, 2025

Uh oh!

Sam-Armstrong Jul 21, 2025

Uh oh!

kallal79 Jul 21, 2025

Uh oh!

Sam-Armstrong Jul 21, 2025

Uh oh!

kallal79 Jul 21, 2025

Uh oh!

Sam-Armstrong Jul 21, 2025

Uh oh!

kallal79 commented Jul 21, 2025 •

edited

Loading

Uh oh!

Sam-Armstrong left a comment

Uh oh!

kallal79 commented Jul 22, 2025 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

kallal79 commented Jul 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Description

Changes Made

JAX Array Frontend (ivy/functional/frontends/jax/array.py)

TensorFlow EagerTensor Frontend (ivy/functional/frontends/tensorflow/tensor.py)

Test Coverage

Testing

Consistency Check

Why This Matters

Checklist

Related Issues

Uh oh!

kallal79 commented Jul 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Test Status Update

Failing Test (Unrelated)

Analysis

Our Changes Are Safe

Our Implementation Tests

Documentation

Uh oh!

Sam-Armstrong left a comment

Choose a reason for hiding this comment

Uh oh!

Sam-Armstrong Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

kallal79 Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

Sam-Armstrong Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

kallal79 Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

Sam-Armstrong Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

kallal79 Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

Sam-Armstrong Jul 21, 2025

Choose a reason for hiding this comment

Uh oh!

kallal79 commented Jul 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes Made:

Updated Implementation:

Uh oh!

Sam-Armstrong left a comment

Choose a reason for hiding this comment

Uh oh!

kallal79 commented Jul 22, 2025 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kallal79 commented Jul 21, 2025 •

edited

Loading

JAX Array Frontend (`ivy/functional/frontends/jax/array.py`)

TensorFlow EagerTensor Frontend (`ivy/functional/frontends/tensorflow/tensor.py`)

kallal79 commented Jul 21, 2025 •

edited

Loading

kallal79 commented Jul 21, 2025 •

edited

Loading