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 TensorFlow Lite (tf.lite) API, including the deprecation of tf.lite.Interpreter in Python with a migration path to ai_edge_litert.interpreter, and changes to certain C++ API constants for improved compatibility. Key updates also include runtime support for the bfloat16 data type in the tfl.Cast operation, alongside 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. Undefined symbol while importing tensorflow: This issue describes a bug where importing TensorFlow version 2.20.0 on a RHEL UBI 9.3 image with Python 3.12 results in an ImportError due to an undefined symbol in the _pywrap_tf_session shared object, indicating a possible compatibility or build problem. The user reports that downgrading Python to 3.11 does not resolve the issue, and attempts to reproduce the problem on other environments failed, suggesting the problem may be specific to the custom build or platform configuration.

   • The discussion includes attempts to diagnose the problem by checking shared library dependencies and suggestions to downgrade Python, which did not fix the error. Contributors requested more environment details and installation steps, while the original poster shared their custom build configuration for Power9 architecture, highlighting the complexity of reproducing the issue and the possibility of platform-specific build challenges.
   • Number of comments this week: 6

2. Images returned by tf.image.decode_jpeg have significant differences with images returned by other common loaders.: This issue reports a bug where images decoded using TensorFlow's tf.image.decode_jpeg differ significantly from those loaded by other common image loaders like PIL, Matplotlib, and Keras utilities, despite appearing visually similar. The user highlights that this discrepancy affects workflows such as using tf.keras.utils.image_dataset_from_directory, which may lead to inconsistent results when mixing image loading methods, and provides code and visual evidence to demonstrate the problem.

   • The comments clarify missing import statements and correct errors in the original test code, with the user sharing a Colab notebook for reproducibility. They also confirm testing the issue on TensorFlow versions 2.19 and 2.20, indicating the problem persists across these versions.
   • Number of comments this week: 3

3. Cannot build lite for iOS: This issue describes a build failure when attempting to compile the TensorFlow Lite framework for iOS using Bazel version 7.4.1 on macOS. The error involves a missing LC_UUID load command in the wrapped_clang_pp binary during compilation, which appears to be related to a known Bazel issue, preventing successful building of the TensorFlowLiteC_framework target.

   • The comments confirm that the same error occurs on macOS even when building the core TensorFlow version (not Lite or iOS specific), suggesting the problem is broader and likely linked to the Bazel issue referenced. One user inquired about any updates, indicating ongoing concern but no resolution provided in the discussion.
   • Number of comments this week: 2

4. plot_model rankdir="LR" show empty image: This issue reports a bug where using the plot_model function with the rankdir="LR" option results in an empty saved image, despite the vertical plotting working correctly. The user highlights that horizontal plotting is preferable for viewing skip connections in a UNet model, but currently fails to produce a visible output.

   • The comment explains that the issue is related to the Keras utility tf.keras.utils.plot_model, which is maintained in a different repository, and advises the user to report the problem there for further assistance.
   • Number of comments this week: 1

5. Support python 3.14: This issue requests the addition of support for Python 3.14 in the project, categorizing it as a feature enhancement. It serves as a tracking point for the ongoing efforts to ensure compatibility with the upcoming Python version, referencing a related issue for context.

   • The single comment references another issue number, indicating a connection or follow-up without further discussion or resolution provided in the comment thread.
   • Number of comments this week: 1

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 that the default driver installation reverts to version 550, which is incompatible with their setup, and despite multiple reinstallations and attempts to fix the environment, the TensorRT warning persists, hindering their ability to run GPU-accelerated TensorFlow operations.
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 near 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 type checking, indicating improper handling of large unsigned 64-bit integers.
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 manually 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 kernel for the complex data type is not found, contradicting the official documentation.
5. tf.raw_ops.Unbatch aborts with "Check failed: d < dims()": This issue reports a bug in TensorFlow version 2.17 where the tf.raw_ops.Unbatch operation aborts with a "Check failed: d < dims()" error, causing the program to crash. The problem occurs when using certain inputs on a Linux Ubuntu 20.04.3 LTS system with Python 3.11.8, and the user has provided a minimal code snippet that reproduces the failure consistently.

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: 9

Summarized Issues:

• TensorFlow Bugs in Model Visualization and Tensor Operations: Several issues report bugs affecting TensorFlow's functionality, including an empty image output when plotting a UNet model horizontally with plot_model(rankdir="LR") and a ValueError caused by tensor shape mismatches during broadcasting in a custom LSTM model. These bugs highlight problems in model visualization and tensor shape handling that disrupt expected workflows.
• [issues/102880, issues/103397]

• Python Version Compatibility and Import Errors: There are compatibility issues with newer Python versions, such as an ImportError when importing TensorFlow 2.20.0 on RHEL UBI 9.3 with Python 3.12 due to undefined symbols in shared objects, and ongoing efforts to add support for Python 3.14. These problems indicate challenges in maintaining TensorFlow's compatibility with the latest Python releases and system environments.
• [issues/103033, issues/102890]

• Quantization and Image Decoding Discrepancies: Issues have been raised about discrepancies in TFLite quantization math outputs compared to official results, and significant pixel-level differences in images decoded by TensorFlow's tf.image.decode_jpeg versus other libraries. These discrepancies point to inconsistencies in TensorFlow's quantization implementation and image processing pipelines.
• [issues/102943, issues/102996]

• Security Vulnerabilities and Dependency Updates: A security vulnerability in Keras (CVE-2025-8747) fixed in Keras 3.11 remains unaddressed in TensorFlow 2.20.0, which still installs Keras 3.10.0. This raises concerns about timely updates of TensorFlow dependencies to mitigate known security risks.
• [issues/103066]

• TensorFlow Lite Issue Diagnostics on Android: A problem with TensorFlow Lite in Google Play Services on Android devices requires detailed system and SDK information, along with logs and minimal test cases, to facilitate diagnosis. This reflects ongoing challenges in troubleshooting TensorFlow Lite issues in mobile environments.
• [issues/103263]

• Documentation Improvements for TensorFlow Ops: A documentation bug in tf.raw_ops.IsotonicRegression calls for clearer docstrings explaining type matching rules and limitations to prevent user confusion and errors. This highlights the need for improved documentation to aid correct API usage.
• [issues/103352]

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: 1

Summarized Issues:

• BatchNormalization axis behavior: The issue highlights that TensorFlow's BatchNormalization layer with axis=1 on inputs shaped (n_test, image_channels, image_size, image_size) does not perform per-instance normalization as expected. Instead, it normalizes across the batch, resulting in minimal transformation of the input compared to PyTorch's InstanceNorm2d(channels).
• issues/102870

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: 4

Key Open Pull Requests

1. [WIP] Support Bzlmod: This pull request is a work-in-progress effort to support Bzlmod in the TensorFlow build system, involving breaking the current non-Bzlmod setup, partial migration of dependencies, and adjustments needed for handling multiple Python versions and pip packaging, with the goal of demonstrating the necessary steps for full Bzlmod integration.

• URL: pull/102891

• Merged: No

• Associated Commits: acc4d, ad38a, 7246a

2. [oneDNN] upgrading oneDNN to 3.8.2: This pull request upgrades the oneDNN library from version 3.7 to 3.8.2, incorporating several bug fixes and has been tested on multiple models across various platforms including cascade-lake, sapphire-rapids, and granite-rapids.

• URL: pull/102932

• Merged: No

• Associated Commits: 9c0f7

3. Add missing pybind11_protobuf dependency to XLA workspace: This pull request adds the missing pybind11_protobuf dependency to the XLA workspace configuration in TensorFlow by creating a new repository definition and updating workspace initialization files, thereby resolving build inconsistencies and runtime import errors related to undefined symbols like GenericProtoCast.

• URL: pull/103312

• Merged: No

• Associated Commits: 2d0c4

Other Open Pull Requests

• GPU matrix condition number estimation fix: This pull request introduces a fix for the near-singular matrix check on GPU by implementing a condition number estimation method. It addresses the specific issue reported in the TensorFlow repository to improve numerical stability during GPU computations.
• pull/103324

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: 6

Key Closed Pull Requests

1. Fix b/288437118: Unconditionally apply clip_by_value in saturate_cast…: This pull request addresses issue b/288437118 by modifying the saturate_cast function in TensorFlow to unconditionally apply clip_by_value, ensuring that infinity values are properly clamped to the target data type's range before casting, thereby preventing undefined behavior and adding tests to verify this fix.

• URL: pull/102861

• Merged: No

• Associated Commits: 7aa89

2. Fix plot_model with rankdir='LR' producing empty images: This pull request addresses the issue of plot_model producing empty images when using rankdir='LR' by replacing the pydot.write() method with an explicit create() call and file write, ensuring the correct graphviz program is invoked and all graph attributes, including rankdir, are properly respected.

• URL: pull/102908

• Merged: No

• Associated Commits: f58a2

3. Spam again: Align Python quantization helpers with TF Lite single-rounding path: This pull request provides a Python implementation of TensorFlow Lite's single-rounding quantization helpers to fix intermittent ±1 discrepancies between Python quantization code and TF Lite runtime by aligning with TF Lite 2.20+'s default single-rounding behavior, includes comprehensive tests, documentation, and demo tools, and enables accurate layer-by-layer debugging of quantized neural network inference without breaking existing APIs.

• URL: pull/103028

• Merged: No

• Associated Commits: 0d7d7

Other Closed Pull Requests

• Security vulnerability proof of concept: This pull request demonstrates a proof of concept for CVE-2025-30066, revealing that the trilom/file-changes-action@v1.2.4 workflow is vulnerable to command injection via malicious filenames. This vulnerability can lead to execution of shell commands and exposure of sensitive workflow secrets or configuration data in CI logs.
• pull/103319

• Build dependency fix: This pull request fixes a build inconsistency by adding the missing pybind11_protobuf dependency to the XLA workspace configuration. It ensures proper repository definition and initialization to resolve runtime import failures related to undefined symbols like GenericProtoCast.
• pull/103311

• Test commit: This pull request is a test commit titled "테스트 커밋" that was not merged into the TensorFlow project. It appears to be for testing purposes only without further changes or fixes.
• pull/103391

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.