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 favor of ai_edge_litert.interpreter and changes to certain C++ constants for improved API flexibility. 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. [TYPE:BUG] [SUBTYPE:WINDOWS] [2.20.0] Tensorflow dll load failing: This issue describes a problem where importing TensorFlow after pandas on Windows 11 causes a DLL load failure, which did not occur in TensorFlow version 2.14.0 but appears in versions 2.18.0 through 2.20.0. The user reports that importing TensorFlow first avoids the error, indicating an import order conflict that leads to the failure of the native TensorFlow runtime initialization.

   • The comments include a detailed package version listing, suggestions that the error might be due to missing dependencies or CPU instruction support, and the user clarifying that these common causes do not apply in their case; the discussion concludes with a recommendation to close the issue as a duplicate of a previously reported one.
   • Number of comments this week: 4

2. [STAT:AWAITING RESPONSE] [TYPE:SUPPORT] [COMP:LITE] Quantização de Modelo para Mobile: This issue concerns the quantization of a model to enable it to run efficiently on Android mobile devices using TensorFlow Lite. The user is requesting guidance or support on how to properly quantize their model for mobile deployment.

   • The comments reiterate the need for model quantization to run on Android devices, and the maintainers have requested the user to provide detailed system information, TensorFlow Lite version, and reproducible code to better diagnose the problem, while also advising to keep the discussion consolidated in this issue.
   • Number of comments this week: 3

3. [TYPE:BUG] [COMP:GPU] [2.17] Blackwell (sm_120) GPU graph execution fails with CUDA_ERROR_INVALID_HANDLE while basic TF GPU ops succeed: This issue describes a persistent TensorFlow GPU runtime failure occurring on NVIDIA Blackwell architecture (sm_120, RTX 5080 Laptop) when executing real SavedModel graph inference, despite the GPU being visible and basic GPU operations succeeding. The failure manifests as a CUDA_ERROR_INVALID_HANDLE during graph execution involving TensorFlow Text and ragged tensor operations, with CPU fallback working only after reloading the model, and attempts to build or use different TensorFlow/tf-text versions revealing compatibility and tooling challenges specific to Blackwell GPUs.

   • The comments detail extensive repro and environment testing confirming the issue is not limited to a single TensorFlow build or packaging problem, highlight a hybrid CPU/GPU workaround isolating the failure to postprocessing steps, and document attempts to disable XLA autotuning and use nightly builds, all of which fail to provide a stable GPU inference path on Blackwell, leaving CPU fallback as the only reliable solution so far.
   • Number of comments this week: 3

4. [TYPE:DOCS-BUG] [AWAITING PR MERGE] [2.21.0] www.tensorflow.org/install outdated (Python 3.13 is supported by pip): This issue reports that the official TensorFlow installation documentation is outdated because it lists Python support only up to version 3.12, whereas the pip package already supports Python 3.13. The user highlights that installing TensorFlow with Python 3.13 works without warnings, indicating the need to update the website to reflect this compatibility.

   • The first comment acknowledges the issue and confirms a pull request has been submitted to update the documentation, while the second comment expresses interest in contributing to the fix.
   • Number of comments this week: 2

5. [TYPE:BUG] [COMP:XLA] [2.21.0] XLA silently accepts out-of-bounds TensorArray.read after TensorArray.unstack, while eager execution raises OutOfRangeError: This issue describes a bug where TensorFlow's XLA compiler silently accepts out-of-bounds reads from a TensorArray after unstacking, whereas eager execution correctly raises an OutOfRangeError for the same operation. This discrepancy leads to inconsistent behavior between eager mode and XLA compiled execution, potentially causing unnoticed errors in compiled TensorFlow functions.

   • The comments confirm the issue was reproducible on Colab with TensorFlow versions 2.21.0 and nightly builds, and the interaction ends with an acknowledgment of the reproduction.
   • 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.

As of our latest update, there are no stale issues for the project this week.

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: 23

Summarized Issues:

• XLA Compilation and Execution Crashes: Multiple issues report that enabling XLA compilation in TensorFlow leads to crashes or segmentation faults in various operations such as GPU execution on NVIDIA Blackwell GPUs, slicing with negative indices, tensor rank checks, and TensorArray operations. These bugs cause inconsistent behavior between eager and XLA modes, often resulting in fatal errors instead of proper exceptions.
• issues/111958, issues/112068, issues/112132, issues/112135, issues/112186, issues/112189

• GPU Runtime Failures on NVIDIA Blackwell GPUs: There are persistent GPU runtime failures on NVIDIA Blackwell (sm_120) GPUs involving CUDA errors and incompatibilities with LLVM PTX versions, causing crashes during inference and SavedModel graph execution despite successful basic GPU operations. Attempts to fallback to CPU or change TensorFlow and tf-text versions have not resolved these issues.
• issues/111958, issues/112272

• Out-of-Bounds Index Handling Inconsistencies: TensorFlow operations such as tf.gather_nd and tf.unravel_index behave inconsistently with expected error handling by allowing out-of-bounds indices without raising errors under certain modes, differing from eager execution and NumPy's documented behavior. This leads to silent acceptance of invalid inputs and inconsistent results.
• issues/112068, issues/112130

• Segmentation Faults and Crashes in TensorFlow Core: Several issues describe segmentation faults and crashes triggered by specific TensorFlow operations or configurations, including boolean tensor concatenation with oneDNN optimizations, setting internal session attributes to None, and invalid intra_op_parallelism_threads values. These faults cause fatal errors instead of proper Python exceptions or error messages.
• issues/112139, issues/112404, issues/112405

• Incorrect or Unexpected Function Behavior: Some TensorFlow functions produce incorrect results or behave unexpectedly, such as tf.signal.kaiser_window producing wrong outputs for negative beta values, tf.raw_ops.Round returning zero for integer inputs, and tf.pad showing random behavior with certain padding modes on large tensors. These issues contradict documented or mathematically expected behavior.
• issues/112200, issues/112212, issues/112304

• Inconsistent Behavior Between Eager and Graph Modes: TensorFlow functions like tf.nn.conv1d exhibit inconsistent behavior where eager execution silently produces zero spatial dimension outputs, but graph mode raises errors during shape inference. This discrepancy causes confusion as code that runs in eager mode fails when wrapped in tf.function.
• issues/112371

• Installation and Environment Issues: Installation instructions are outdated regarding Python version support, and environment-specific problems such as DLL load failures on Windows 11 due to import order and macOS build failures related to DYLD_LIBRARY_PATH cause runtime errors and build failures.
• issues/112027, issues/112207, issues/112309

• TensorFlow API Crashes with Specific Inputs: Certain TensorFlow APIs crash unexpectedly when given specific inputs, such as tf.compat.v1.spectral.irfft crashing on complex64 tensors of shape (1,), and tf.nn.ctc_loss crashing when logits have zero classes. These crashes occur without proper error handling.
• issues/112406, issues/112407

• Model Quantization and Publishing Limitations: There are issues related to model quantization for efficient Android deployment and limitations in TensorFlow Maven publishing that prevent app publication due to unsupported request sizes.
• issues/111989, issues/112171

• Placeholder and Testing Issues: A placeholder issue was created solely to test the project environment setup and does not describe a functional problem.
• issues/112345

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: 5

Summarized Issues:

• Environment Setup Placeholders: Several issues serve as placeholders to assist new contributors with environment setup, providing a structured way to onboard beginners. These issues are sequentially numbered and focus on easing the initial setup process for the TensorFlow project.
• issues/112346, issues/112347, issues/112348, issues/112349

• Quantization Topic: One issue addresses a topic related to quantization, referencing a specific comment on a related TensorFlow GitHub issue. This indicates ongoing discussion or investigation into quantization within the project.
• issues/112174

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: 28

Key Open Pull Requests

1. Fix logdet general matrices: This pull request fixes the tf.linalg.logdet function to correctly compute the logarithm of the determinant for general square matrices by switching from Cholesky decomposition to LU decomposition, thereby supporting non-symmetric positive definite matrices with positive determinants and aligning its behavior with PyTorch and NumPy, while also addressing several related issues.

• URL: pull/112118

• Associated Commits: 05271, 5ff47, c5858, b5601

2. chore: Migrate gsutil usage to gcloud storage: This pull request automates the migration of all Google Cloud Storage command-line interactions in the project from the legacy gsutil tool to the modern gcloud storage CLI, aiming to improve performance, unify authentication, and standardize command usage while requiring thorough review and testing to address any differences in command behavior or output.

• URL: pull/112004

• Associated Commits: c4c0c, c26c4

3. Fix Windows GPU warning triggering on internal Keras calls: This pull request fixes the issue where the Windows GPU deprecation warning was incorrectly triggered during internal Keras calls to list_physical_devices() in model.compile(), by adding a check to ensure the warning only appears when the user explicitly queries for GPU devices.

• URL: pull/112008

• Associated Commits: 4edec, 30b37

Other Open Pull Requests

• Stale Issue Workflow Updates: Multiple pull requests increase the inactivity period before automatically closing stale issues from 30 days to 60 or 90 days to provide contributors with more time to address concerns and better align with the project's activity level. These changes aim to reduce premature issue closures and encourage more community engagement by extending the time inactive issues remain open.
• [pull/112315, pull/112294, pull/112323, pull/112344, pull/112365, pull/112368, pull/112376, pull/112384, pull/112403, pull/112421, pull/112426, pull/112428, pull/112451]

• Fixes to Mathematical and Kernel Operations: Several pull requests address fixes in mathematical functions and GPU kernel loops, including handling complex reciprocal calculations with infinite values, fixing tf.math.polyval behavior for single-coefficient polynomials, correcting tf.linalg.logdet to handle non-symmetric positive-definite matrices by switching to LU decomposition, and resolving integer overflow in GPU kernel loops by supporting 64-bit indices. These improvements ensure correctness and stability in TensorFlow's numerical computations and GPU operations.
• [pull/111977, pull/111980, pull/112110, pull/112249]

• Documentation and Warning Enhancements: Some pull requests improve documentation by adding detailed warnings about security risks in tf.image.resize, updating the add_metric() docstring to reflect Keras 3 changes, fixing broken padding documentation links, and correcting grammatical errors in the README installation section. These updates clarify usage, raise awareness of potential issues, and improve overall documentation quality.
• [pull/111981, pull/111982, pull/112288, pull/112422]

• Stale Issue Workflow Parameter Adjustments: A subset of pull requests specifically adjust parameters in the stale issue workflow, such as increasing days-before-close and days-since-last-update from 30 to 60 days, or reducing days-before-close from 90 to 30 days, to fine-tune issue management timing. These changes help balance timely issue resolution with adequate time for community response.
• [pull/112294, pull/112323, pull/112344, pull/112368, pull/112376, pull/112384, pull/112403, pull/112426, pull/112428, pull/112451, pull/112422]

• XLA and Backend Fixes: Pull requests include raising the minimum PTX version to 8.7 for Blackwell GPUs to prevent JIT compilation crashes and fixing a crash in the XLA StridedSlice operation caused by degenerate dynamic ranges by clamping dimension sizes. These backend fixes improve stability and compatibility of TensorFlow on specific hardware and edge cases.
• [pull/112039, pull/112424]

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: 0

As of our latest update, there are no closed pull requests for the project this week.

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.