Weekly GitHub Report for Tensorflow: August 04, 2025 - August 11, 2025 (22:42:03)

Weekly GitHub Report for Tensorflow

Thank you for subscribing to our weekly newsletter! Each week, we deliver a comprehensive summary of your GitHub project's latest activity right to your inbox, including an overview of your project's issues, pull requests, contributors, and commit activity.

---

Table of Contents

• I. News
  • 1.1. Recent Version Releases
  • 1.2. Other Noteworthy Updates
• II. Issues
  • 2.1. Top 5 Active Issues
  • 2.2. Top 5 Stale Issues
  • 2.3. Open Issues
  • 2.4. Closed Issues
  • 2.5. Issue Discussion Insights

• III. Pull Requests
  • 3.1. Open Pull Requests
  • 3.2. Closed Pull Requests
  • 3.3. Pull Request Discussion Insights

• IV. Contributors
  • 4.1. Contributors

I. News

1.1 Recent Version Releases:

The current version of this repository is v2.19.0

1.2 Version Information:

Released on March 5, 2025, TensorFlow version 2.19.0 introduces breaking changes to the tf.lite API, including the deprecation of tf.lite.Interpreter in favor of ai_edge_litert.interpreter and changes to certain C++ constants for improved API compatibility. Key updates also include runtime support for the bfloat16 data type in the tfl.Cast operation and the discontinuation of standalone libtensorflow package publishing, while still allowing unpacking from PyPI.

II. Issues

2.1 Top 5 Active Issues:

We consider active issues to be issues that that have been commented on most frequently within the last week. Bot comments are omitted.

1. Importing TF 2.20.0rc0 and PyArrow results in a Abort trap: 6 on MacOS ARM: This issue reports a bug where importing TensorFlow 2.20.0rc0 together with PyArrow on MacOS ARM results in a crash with an abort trap due to a mutex lock failure. The problem appears related to a conflict between different protobuf versions used by TensorFlow and PyArrow, causing deadlocks or import errors depending on the import order and protobuf version compatibility.

   • The discussion confirms the issue occurs on ARM CPUs and involves an ODR violation likely caused by protobuf version mismatches between TensorFlow 2.20.0rc0 and PyArrow 21.0.0. Attempts to pin protobuf versions reveal incompatibilities, and similar crashes are reported when using other packages like ray[tune] that depend on PyArrow, with the root cause traced back to PyArrow’s protobuf dependency.
   • Number of comments this week: 9

2. tf.linalg.lu_solve returns inconsistent output from CPU vs GPU: This issue reports a bug where the tf.linalg.lu_solve function returns significantly different results when run on CPU versus GPU using float64 tensors with large values in TensorFlow versions 2.19.0 and 2.21.0-dev20250801. The user provides reproducible code demonstrating that the CPU output is near zero while the GPU output is a large negative number, indicating inconsistent behavior between devices.

   • The commenters attempted to reproduce the issue locally without success, suggesting the problem might be environment-specific. They recommended verifying GPU availability in TensorFlow to ensure the GPU is properly accessed, as the discrepancy might be due to both runs occurring on the CPU rather than a true CPU-GPU inconsistency.
   • Number of comments this week: 3

3. Inconsistent output after model reload when using @tf.function on call: This issue reports inconsistent outputs from a custom Keras model’s call() method when decorated with @tf.function after saving and reloading the model, causing assertion failures due to output divergence. The problem disappears when @tf.function is removed, indicating a potential bug related to graph tracing or serialization in TensorFlow 2.19.0 on Ubuntu with a specific hardware and software setup.

   • The first comment confirms the issue can be reproduced and notes that it does not occur when using the new Keras 3 API with the .keras format, providing a Colab example for reference. The second comment asks whether the older TensorFlow 2/Keras save formats will be fixed or if users should migrate to Keras 3, seeking official guidance on future support and recommendations for migration strategies.
   • Number of comments this week: 2

4. Invalid class_weight silently ignored in model.fit(), no error raised: This issue reports a bug where passing an invalid class_weight dictionary with non-integer keys to model.fit() in TensorFlow does not raise any error or warning, causing the training to proceed silently without applying the intended class weights. The expected behavior is for the function to raise a ValueError or KeyError when invalid keys are provided, but currently, the invalid input is ignored without notification, potentially leading to unnoticed misbehavior during model training.

   • The comments indicate that a pull request has been submitted to fix the issue, and the maintainers acknowledge the contribution and express gratitude to the user for reporting the problem, assuring that the issue will be resolved once the PR is merged.
   • Number of comments this week: 2

5. tf.math.abs returns inconsistent results for complex inf/nan inputs on CPU and GPU: This issue reports a bug where the tf.math.abs function returns inconsistent results on CPU and GPU when given complex inputs containing both infinite and NaN values, specifically showing [inf] on CPU and [nan] on GPU. The user has provided reproducible code demonstrating this discrepancy in TensorFlow version 2.17 and confirmed the bug with TensorFlow Nightly builds.

   • A pull request has been created to address the issue, and community members have acknowledged the submission, indicating that once the PR is merged, the problem should be resolved.
   • Number of comments this week: 2

2.2 Top 5 Stale Issues:

We consider stale issues to be issues that has had no activity within the last 30 days. The team should work together to get these issues resolved and closed as soon as possible.

1. TF-TRT Warning: Could not find TensorRT: This issue describes a problem where TensorFlow on Ubuntu 22.04 cannot detect TensorRT despite having an NVIDIA RTX 3050 Ti GPU with the appropriate CUDA 12.4 and NVIDIA driver 535 installed. The user reports difficulties with driver compatibility and repeated installation attempts, resulting in a persistent warning "TF-TRT Warning: Could not find TensorRT," which is hindering their ability to proceed with machine learning development.
2. SystemError in tf.ensure_shape and tf.compat.v1.ensure_shape when dtype of shape is tf.uint64 and its value is too large.: This issue reports a bug in TensorFlow where using tf.ensure_shape or tf.compat.v1.ensure_shape with a shape tensor of type tf.uint64 containing very large values close to 2^64 causes a SystemError and OverflowError. Specifically, when such large uint64 values are passed in eager execution mode, the functions fail with an internal error related to the built-in isinstance function, indicating improper handling of large unsigned integer shapes.
3. Feature Request: Integrate different Digital Signal Processing into tf.signal: This issue is a feature request proposing the integration of advanced digital signal processing (DSP) functionalities, similar to those found in the PyTorch julius library, into TensorFlow's tf.signal module. The requester highlights the benefits of having native audio data augmentation tools within TensorFlow to improve model robustness and streamline workflows by reducing reliance on external libraries.
4. [DOCS] Missing complex input for Round op: This issue reports a documentation bug in TensorFlow where the Round operation is described as supporting complex tensor inputs, but in practice, attempting to use a complex tensor with this operation results in an error, requiring users to round the real and imaginary parts separately. The user provides a reproducible example on MacOS with TensorFlow 2.15.0 and highlights a device-related error indicating that the Round op does not properly handle complex inputs as documented.
5. tf.raw_ops.Unbatch aborts with "Check failed: d < dims()": This issue reports a bug in TensorFlow version 2.17 where the operation tf.raw_ops.Unbatch aborts with a fatal check failure error "Check failed: d < dims()" when invoked with certain tensor inputs. The user has reproduced the problem on Linux Ubuntu 20.04.3 using Python 3.11.8 and provided a minimal code snippet that triggers the crash, indicating a potential flaw in the handling of tensor dimensions within the Unbatch operation.

2.3 Open Issues

This section lists, groups, and then summarizes issues that were created within the last week in the repository.

Issues Opened This Week: 12

Summarized Issues:

• Inconsistent numerical results between CPU and GPU: Several issues report discrepancies in TensorFlow operations producing different or inconsistent outputs when run on CPU versus GPU. These include tf.linalg.lu_solve with large float64 tensors and tf.math.multiply and tf.math.abs functions with complex tensors containing infinity and NaN values, indicating hardware-specific numerical inconsistencies.
• [issues/98345, issues/98408, issues/98410]

• Crashes and aborts on GPU operations: Using certain TensorFlow raw operations on GPU causes crashes and abort signals, while the same operations run successfully on CPU. Specifically, tf.raw_ops.MatrixDiag with a 6-dimensional tensor input triggers a GPU crash with a core dump.
• [issues/98349]

• Build and compatibility failures on specific platforms: Multiple issues describe build failures or runtime errors due to platform-specific incompatibilities. These include a macOS 13.6 build failure requiring Xcode version specification, a Windows 10 binary incompatibility caused by mismatched NumPy versions, and a Windows 11 DLL load error preventing TensorFlow initialization.
• [issues/98361, issues/98406, issues/98640]

• Library import conflicts causing crashes: Importing TensorFlow alongside other libraries can cause crashes due to conflicting dependencies. For example, on MacOS ARM, importing TensorFlow 2.20.0rc0 with PyArrow 21.0.0 results in an abort trap caused by mutex lock failures likely due to incompatible protobuf versions.
• [issues/98563]

• TensorFlow Lite and model conversion inefficiencies: Issues with TensorFlow Lite include unsupported layer conversions and alignment requirements. A customized QAT model with an unsupported Multiply layer prevents BatchNorm fusion during TFLite conversion, increasing inference time, and the TensorFlow Lite task audio library for Android lacks required 16KB page alignment, causing build warnings and uncertainty about future support.
• [issues/98324, issues/98489]

• Silent failure on invalid input parameters: Passing an invalid class_weight dictionary with non-integer keys to model.fit() in TensorFlow 2.19.0 does not raise any error or warning, causing the training to silently ignore the invalid weights instead of raising a ValueError or KeyError as expected.
• [issues/98283]

• Loss subclass API causing poor model convergence: Implementing a custom contrastive loss using the Loss subclass API in TensorFlow 2.19.0 results in poor model convergence compared to using an equivalent wrapper function. This regression was not present in TensorFlow 2.15.0, suggesting a potential internal bug in the subclass API for loss functions.
• [issues/98477]

2.4 Closed Issues

This section lists, groups, and then summarizes issues that were closed within the last week in the repository. This section also links the associated pull requests if applicable.

Issues Closed This Week: 2

Summarized Issues:

• DLL Load Errors on TensorFlow Import: Multiple issues report failures when importing TensorFlow modules due to DLL load errors related to the native TensorFlow runtime. These errors occur across different TensorFlow versions and Python versions, causing ImportErrors during initialization of internal components.
• issues/98359, issues/98488

2.5 Issue Discussion Insights

This section will analyze the tone and sentiment of discussions within this project's open and closed issues that occurred within the past week. It aims to identify potentially heated exchanges and to maintain a constructive project environment.

Based on our analysis, there are no instances of toxic discussions in the project's open or closed issues from the past week.

---

III. Pull Requests

3.1 Open Pull Requests

This section provides a summary of pull requests that were opened in the repository over the past week. The top three pull requests with the highest number of commits are highlighted as 'key' pull requests. Other pull requests are grouped based on similar characteristics for easier analysis. Up to 25 pull requests are displayed in this section, while any remaining pull requests beyond this limit are omitted for brevity.

Pull Requests Opened This Week: 6

Key Open Pull Requests

1. Add validation for class_weight keys in standardize_sample_or_class_weight.: This pull request adds comprehensive validation for class_weight keys in the standardize_sample_or_class_weight function and the data adapter layer within TensorFlow's Keras engine, ensuring invalid keys are properly rejected with detailed error messages while preserving sample_weight functionality, and includes a new test suite covering various validation scenarios for both single-output and multi-output models.

• URL: pull/98429

• Merged: No

• Associated Commits: 65695, 6f481, 9e0f3, 3bb69, 5dd18, 3029f, 1bc39

2. tf.math.abs returns inconsistent results for complex inf/nan inputs on CPU and GPU fix: This pull request addresses issue #98410 by adding IEEE 754 compliant template specializations for complex64 and complex128 types in GPU kernels to ensure that tf.math.abs returns consistent results, specifically returning infinity instead of NaN when either the real or imaginary component is infinite, thereby fixing discrepancies between CPU and GPU outputs.

• URL: pull/98518

• Merged: No

• Associated Commits: 31bd1, b21ca

3. FIX(tensorflow-lite) system path is prefered over TFLITE_HOST_TOOLS_DIR: This pull request modifies the TensorFlow Lite CMake configuration to change the find_program command from using PATHS to HINT, ensuring that the flatc binary located in the TFLITE_HOST_TOOLS_DIR is preferred over the system PATH environment variable during cross-compilation.

• URL: pull/98643

• Merged: No

• Associated Commits: 2987b

Other Open Pull Requests

• Logic error fix in _none_to_default function: This pull request corrects the behavior of the _none_to_default function by ensuring it returns the actual input value when it is not None, and only returns the default value when the input is None. This fix addresses a logic error that previously caused incorrect return values.
• pull/98644

• Grammar improvements in error messages: These pull requests enhance the clarity and correctness of error messages by fixing grammatical errors. One changes "is not a property value" to "is not a valid property value," and another improves the regex_replace function docstring by correcting a phrase to "which can be used to insert."
• pull/98645, pull/98646

3.2 Closed Pull Requests

This section provides a summary of pull requests that were closed in the repository over the past week. The top three pull requests with the highest number of commits are highlighted as 'key' pull requests. Other pull requests are grouped based on similar characteristics for easier analysis. Up to 25 pull requests are displayed in this section, while any remaining pull requests beyond this limit are omitted for brevity.

Pull Requests Closed This Week: 9

Key Closed Pull Requests

1. Dependabot/docker/tensorflow/tools/tf sig build dockerfiles/ubuntu ed1544e: This pull request updates the github/codeql-action GitHub Action from version 3.29.4 to 3.29.5 to incorporate the latest improvements and fixes in the CodeQL CLI and language packs.

• URL: pull/98585

• Merged: No

• Associated Commits: 044fb, f1179, cc6b6, 138ab, dc631, 26372, 00154, 2c0aa, a5f07, b9599, 775cf, 4486b, 3af8b, 6fc68, edf1f, 015ee, 949ab, 7c8a6, 42128, 1521b, 52605, 3e006, b67de, 7ea43, 25aff, ebac8, 6112e, eb67f, 79e7a, 5cb04, 12958, 9e6a0, ce916, 36df3, 87c40, b8b35, 72ae5, f14d6, c1c1d, 541fd, b9c26, a467b, 02402, 093ba, 1dafe, 00db7, df293, 9da35, 02058, 6bfc9, 56829, 31c00, 81139, a9bd1, 5406a, ade7d, 260ab, 7c55e, 8d747, 1aa7a, 266c7, 0d3ee, 8170e, 3e123, 9ce55, f11b0, a8760, c6622, 4c69e, 0ec46, 74c18, 16902, fee16, dc3a1, 2c722, 15e1a, 4d5da, d6e31, 776f6, ddc5b, 2c42f, 4d3f6, 33293, 1415b, 7ba73, 2e553, 75075, 3e1e2, cebb8, 2d15f, 4928e, 6d2c2, d5ec0, 95c69, 63e4c, b9bd8, 3a62c, db602, 16fb3, a683a, 7afe7, cf414, 944a4, e795b, 1f2c6, d778d, 75beb, 1afca, 068eb, 75a52, 0c9ed, 4e32b, 31d70, cc4e5, 50e8c, 765b8, 6dd87, 371d5, aa65f, fec12, dc4cd, 8238d, 49368, d3be2, d8f58, 14c5b, 99583, 26f7a, b5aa0, b3d12, bf0a6, fb9e9, 6317a, 9954f, 6d271, b238e, 46216, bcc2e, 9ebfe, e8ac6, 87944, 5e21a, c7d25, a97ca, fdcfe, db8a3, f44d3, 53db5, a6ffc, 17020, b9445, 658d5, 126ba, 599b9, 81b61, 60a97, 4d2ed, 66a00, e5ce8, da686, 5f4ac, 5e766, a65a1, 8ed65, 4afda, 08334, 76df7, 9d919, 1b78d, 0db35, 5b351, fec2b, dab88, 0acfd, 2ee18, a1e7a, 72b08, ee218, d1f63, a8237, ac292, 8bd27, 4d7be, 8b4e0, 52dd7, 418ab, 7fe14, 87c6c, beaa2, 938ac, d485d, 95485, 7a4ac, 8f8fd, cf252, 371da, 668e3, d5f29, 9e60b, 619cb, 29c3a, 94246, ad74a, 93010, d74b3, 85326, 320c6, d9a0e, d9d6a, 836f4, d093a, 3b3ff, faea1, 8e10b, bcc38, 9211c, ddc15, c7343, 8cfda, ac7c9, 7bbac, 79f01, 66125, 221f2, 842b5, fdc38, c779d, bad19, 9ffef, 61fde, b7094, 5aaff, 58130, 80a09, 2de52, cb9e9, 1f58d, ceb8e, 95e3c, c97ab, 36a33, cd122, a8051, f4d00, 53eee, d41c5, 5f2dc, 92e0c, a1b03, 7f82f

2. Update RELEASE.md with 2.19.1 release notes: This pull request updates the RELEASE.md file with the release notes for version 2.19.1, including a minor fix to remove certain notes.

• URL: pull/98439

• Merged: 2025-08-07T04:58:38Z

• Associated Commits: 921c5, 23b92

3. Update release notes for TensorFlow 2.19.1: This pull request is an incomplete update to the release notes for TensorFlow version 2.19.1, intended for a Release Owner to fill in the internal release notes before submission.

• URL: pull/97297

• Merged: No

• Associated Commits: 64045

Other Closed Pull Requests

• Modernizing tf_env_collect.sh string formatting: Multiple pull requests focus on updating the tf_env_collect.sh script by replacing traditional %-based string formatting with Python f-strings. These changes aim to improve code readability, maintainability, and performance without altering functionality, although some were not merged.
• pull/98397, pull/98399, pull/98400

• TensorFlow version number updates: Several pull requests update TensorFlow version numbers in key files to reflect new patch and minor releases, such as 2.19.1 and 2.20.0. They also include warnings about ensuring no old version strings remain in the source directory before merging.
• pull/98437, pull/98438

• Dependency upgrade for h5py: One pull request upgrades the h5py version in requirement_lock files by cherry-picking a specific commit to maintain dependency consistency.
• pull/98583

3.3 Pull Request Discussion Insights

This section will analyze the tone and sentiment of discussions within this project's open and closed pull requests that occurred within the past week. It aims to identify potentially heated exchanges and to maintain a constructive project environment.

Based on our analysis, there are no instances of toxic discussions in the project's open or closed pull requests from the past week.

---

IV. Contributors

4.1 Contributors

Active Contributors:

We consider an active contributor in this project to be any contributor who has made at least 1 commit, opened at least 1 issue, created at least 1 pull request, or made more than 2 comments in the last month.

If there are more than 10 active contributors, the list is truncated to the top 10 based on contribution metrics for better clarity.

Contributor	Commits	Pull Requests	Issues	Comments
tensorflower-gardener	185	0	0	0
mihaimaruseac	0	0	0	66
harshithn31	43	0	0	0
ezhulenev	35	0	0	0
djxnqp	0	0	31	0
Aaraviitkgp	10	3	0	7
No author found	14	0	0	0
WillFroom	13	0	0	0
Venkat6871	1	1	0	10
beckerhe	12	0	0	0

Access Last Week's Newsletter:

• Link